Fast Ion Tails during Radio Frequency Heating
on the Alcator C-Mod Tokamak
by
Jon Christian Rost
B.S., Physics
B.S., Electrical Engineering
University of Maryland, College Park, 1991
Submitted to the Department of Physics in Partial
Fulfillment of the Requirements for the Degree of
Doctor of Philosophy in Physics
at the
Massachusetts Institute of Technology
June 1998
Copyright 1998 Massachusetts Institute of Technology
All rights reserved
Signature of Author:
a
oDepartment
of Physics
April 10, 1998
Certified by: _
Miklos Porkolab
Professor of Physics
Thesis Supervisor
Accepted by:
C~i", :..,,l
JUN 091998
LIBRARIES
Thomas J. Greytak
Associ ttepartment Head for Education
Fast Ion Tails during Radio Frequency Heating
on the Alcator C-Mod Tokamak
by
Jon Christian Rost
Submitted to the Department of Physics
on April 10, 1998 in Partial Fulfillment of the
Requirements for the Degree of
Doctor of Philosophy in Physics
Abstract
Observations of ion tails in the plasma edge during radio frequency heating on the
Alcator C-Mod tokamak have been made using a toroidally and poloidally scanning
charge-exchange neutral particle analyzer. The ion tails create a large flux of chargeexchange neutrals (hydrogen and deuterium) at suprathermal energies. The neutral
particle flux is characterized by: fast rise and decay times, much faster than the time
for changes in the bulk plasma; dependence on plasma conditions and magnetic field
in the scrape-off layer; a threshold in electric field near the antenna; and no correlation
with bulk plasma parameters. During hydrogen minority heating, edge ion heating
may occur at power levels above approximately 500 kW. When the heating frequency
is at an ion cyclotron harmonic in the edge, edge heating occurs at power levels lower
than 10 kW. Dedicated experiments showed that edge ion heating does not generate
impurities or cause loss of heating efficiency. Using the pitch-angle dependence of
the fast particles, the total parasitic power loss to the edge is estimated at less than
0.1% during standard heating schemes, but up to 5% with a cyclotron resonance in
the edge.
Evidence of Parametric Decay Instability (PDI) into an ion Bernstein wave and
an ion cyclotron quasimode has been found on C-Mod using RF probes. Calculations
of convective thresholds for PDI have been made for a range of edge parameters and
magnetic fields. Calculated theoretical thresholds for PDI in the antenna near field
are consistent with experimentally observed thresholds for edge heating, and observed
dependence on toroidal field and changes at H-mode transitions.
Thesis Supervisor: Miklos Porkolab
Title: Professor of Physics
Acknowledgements
I would first like to thank Professor Miklos Porkolab, who was my research advisor for
four years. His classes in plasma physics formed the foundation of my knowledge of
waves in plasmas. Most importantly, he interceded at a difficult time for me, probably
saving my graduate career and giving me the opportunity to work with him and the
Alcator Group.
Dr. Rejean Boivin worked closely with me throughout my time on Alcator. He
had nearly completed the PCX diagnostic when I arrived at Alcator, and all of the
data used in this thesis was acquired and analyzed with his help. He taught me a lot
about how to be an effective experimental physicist. R6jean also spent uncountable
hours working with me on my thesis, and helping me organize posters and talks. He
was the one who was there on a day-to-day basis to discuss the results from the most
recent data, analysis, or calculation, or just my problems navigating the waters of
graduate school.
This thesis would not exist today if I had not had the help of these two men.
The RF probe data were taken with probes built and operated by my fellow
graduate student and good friend, James Christian Reardon.
The intelligence and hard work of the members of the Alcator Group never ceased
to amaze me. I think Alcator will forever define for me the standard for how physics
should be done. I also appreciate the unique level of support Alcator gives to graduate
students in machine time and equipment.
The advice and enthusiasm of Dr. Robert Pinsker helped immensely in planning,
running, and interpreting these experiments.
I cannot begin to list the other friends who have been with me during my time
through graduate school, giving me a life outside the lab.
I must finally thank my parents who have always supported my decisions about
what path to take in life.
Contents
Abstract
Acknowledgements
1 Introduction
1.1 Ion Cyclotron Resonance Frequency Heating on Tokamaks
1.2 The Alcator C-Mod Neutral Particle Analyzer
1.3 Edge Ion Heating . . . . . . . . . ..
1.4 Parametric Decay Instability .....
1.5 Outline of this Work . . . . . . ...
I
I
II II
2 Plasma operation on Alcator C-Mod
2.1 Sample Alcator C-Mod Shot.....
2.2 Radio-frequency Heating ........
2.2.1 Heating Schemes . . . . . . .
2.2.2 Antenna . . . . . . . . . . . .
2.2.3 RF Fields in the Plasma Edge
2.3 Diagnostics . . . . . . . . . . . . . .
2.3.1 General diagnostics . . . . . .
2.3.2 RF Probes . . . . . . . ....
2.3.3 Impurity Measurements . . .
2.3.4 Neutral Density ........
3 Perpendicular Charge Exchange Analyzer
3.1 Source of CX Neutrals .........................
3.1.1 Source Rate . . . . . . . . . . . . .
3.1.2 Attenuation . . . . . . . . . . . . .
3.2 Apparatus.........................................
3.2.1 Beam Line . . . . . . . . . . . . . .
3.2.2 Stripping Cell . . . . . . . . . . . .
3.2.3 M ain Chamber .........................
3.2.4 M C P 's . . . . . . . . . . . . . . . .
3.2.5 Electronics . . . . . . . . . . . . . .
3.2.6 Movem ent . . . . . . .. .. . . . .
3.3 Absolute Calibration ..........................
: : : : : I :
. . . . . . . . . . . . . .
. . . . . . . .. . . . . .
. . . . . . . . . . . . . .
. . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . .
3.4
3.5
Sources of Error . . .
. . . . . . . .
3.4.1 Counting Statistics ........................
3.4.2 Pick-up . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.4.3 Systematic Error .........................
Sample Data .. .. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4 Dynamics of Edge Ions
4.1
4.2
Behavior of Fast Ions in Tokamaks . . .
4.1.1 Tokamak Magnetic Fields . . . .
4.1.2 Ion Orbits . ......................
Confinement of Edge Ions ........................
4.2.1 Unconfined Orbits ........................
4.2.2 Collisions . . . . . . . . . . . . .
4.2.3 Charge Exchange .........................
4.2.4 D iscussion . . . . . . . . . . . . .
43
43
43
44
45
51
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
....
... .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
51
52
53
60
60
61
62
62
5 Previous Observations of Edge Ion Heating
65
6
71
71
Observations of Edge Ion Heating on Alcator C-Mod
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.2
Identification as Edge Heating . . . . . . . . . .
6.2.1 Tim e Scale . . . . . . . . . . . . . . . .
6.2.2 Tail Temperature . . . . . . . ..... ..........
6.2.3 Dependence on Edge Conditions . . . . .
6.3 Edge Heating during H Minority Heating . . . .
6.3.1 Scatter Plots . . . . . . . . . . . . . . .
6.4 Resonance at Plasma Edge ........................
6.4.1 Toroidal Field Ramps .......................
6.4.2 Scans in RF Power ........................
6.5 ICRF Heating at 6.5 T ..........................
6.6 Im purities . . . . . . . . . . . . . . . . . . . . .
6.6.1 Toroidal Field Ramps .......................
6.6.2 R esults . . . . . . . . . . . . . . . . . . .
6.7 RF Loading and Heating Efficiency . . . . . . .
6.8 H -modes . . . . . . . . . . . . . . . . . . . . . .
6.8.1 Changes in Edge due to H-mode . . . . .
6.8.2 Fast Jump at Transition .....................
6.8.3 Slow Precursor ..........................
6.8.4 Conclusions . . . . . . . . . . . . . .. .
6.9 Power Deposition .............................
6.9.1 M ethods . . . . . . . . . . . . . . . . . .
6.9.2 Resonance in Plasma Edge . . . . . . . .
6.10 Summ ary . . . . . . . . . . . . . . . . . . . . .
. . . . . . .
. . . . . . .
.
. . . . . . .
. . . . . . .
. . . . . . .
. . .
. . .
.......
. . .
. . .
. . .
. .
. .
. .
. .
. .
. . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. . . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . . .
73
73
73
74
78
78
81
82
87
89
89
90
92
92
92
94
94
96
97
97
97
99
101
7 Mechanisms for Generating Energetic
7.1 Damping of the EM Wave ......
7.2 Kinetic Effects .............
7.2.1 Quiver Motion ........
7.2.2 Ponderomotive Force.....
7.3 RF Sheaths ..............
7.4 Electrostatic Modes . . . . .. . . ..
7.4.1 Parametric Decay Instabilities
7.4.2 IBW Launching . . . . . . . .
7.5 Conclusions ..............
Ions in the Plasma Edge
. . . . . . . . . . . . . . . . . .
. .... ... .... .... ..
. .... ... .... .... ..
. .... ... .... .... ..
. . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . .
103
104
104
105
105
106
106
107
107
107
8
Parametric Decay Instabilities
8.1 PDI Theory ..............
.. .... ... .... .... .
8.1.1 Waves .............
... ... .... ... .... .
8.1.2 Dispersion Relation......
... ... .... ... .... .
8.1.3 Growth rate ..........
... .... ... ... .... .
8.1.4 Convection ..........
... .... ... ... .... .
.. . . . . . . . . . . . . . . . . . . .
8.2 Numerical Calculations .....
8.2.1 PDI during Hydrogen Minority Heating . . . . . . . . . . . . .
8.2.2 Resonance in Plasma Edge . . . . . . . . . . . . . . . . . . . .
8.3 RF Probes ...............
... .... ... .... ....
8.3.1 Correspondence with PCX MeaLsurements . . . . . . . . . . .
8.3.2 RF Probe Measurements of PD I . . . . . . . . . . . . . . . . .
8.4 Conclusions ..............
. .. ..... .. .... ....
109
110
110
112
114
115
117
117
122
126
127
128
130
9
Summary and Conclusions
9.1 Effects of Edge Heating .........................
9.2 Parametric Decay Instability as Cause of Edge
9.3 Comparison with Results of other Machines .
9.3.1 Im purities . . . . .
.......
. . . .
9.3.2 Parametric Decay Instabilities . . . . .
9.3.3 Parasitic Power Loss ......................
9.4 Open Questions . . . . . . . . . . . . . . . . .
133
133
134
135
135
136
137
137
Heating
. . . . .
. . . . .
. . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. . . . . . . . . . . .
.
.
.
.
.
.
.
A Magnitude of RF Electric Field in Propagating Fast Wave
139
B Ponderomotive Force in a Magnetized Plasma
143
C Near Field of RF Antenna
C.1 Vacuum Field ...
.. .. ...
.. ..
C.2 Effect of the Plasma ..........................
..
..
...
..
..
..
..
147
. . 147
. 148
10
Chapter 1
Introduction
This thesis describes the exploration of a phenomenon observed during ion cyclotron
resonance frequency (ICRF) heating on the Alcator C-Mod tokamak. Measurements
with a neutral particle analyzer showed the presence of ions with energy far above the
thermal level in the plasma edge during RF heating. These fast ions are an unwanted
side-effect of the RF heating. This work examines the effect the suprathermal edge
ions have on the tokamak plasma and explains their source through the theory of a
non-linear interaction between the plasma and the RF wave.
1.1
Ion Cyclotron Resonance Frequency Heating
on Tokamaks
To the extent the that goal of tokamak research is to achieve a functioning fusion
reactor[1], ion cyclotron resonance frequency heating is a critical part of the field. A
reactor is expected to require a source of heating beyond the resistive heating from
the plasma current, and possibly a means to control the distribution of particles and
current in the plasma[2]. The two dominant methods of auxiliary heating are neutral
beam injection[3] and ICRF heating[2], though electron cyclotron resonance heating
is also being developed. These methods are capable of coupling large amounts of
power to the plasma, and each has its advantages and disadvantages.
On the Alcator C-Mod tokamak, several ICRF heating schemes are employed[4].
The goal of RF heating is to launch an RF wave from outside the plasma which penetrates to the plasma interior and is absorbed on either the ion or electron population.
There is a gap between the RF antenna and the plasma where the electromagnetic
wave is evanescent. To reach the plasma interior, the RF power must cross this gap[5].
Various phenomena in the plasma edge can reduce the RF power to the plasma
interior by absorption in the edge or by decreasing the coupling from the antenna
to the plasma[6]. If RF power is absorbed in the edge, energy from the RF fields is
transferred to ions or electrons. If the coupling is decreased, then more RF power is
reflected from the antenna back into the RF transmission line.
1.2
The Alcator C-Mod Neutral Particle Analyzer
This work centers around measurements made by the neutral particle analyzer (NPA)
on Alcator C-Mod. The neutrals observed result from charge exchange reactions and
actually represent, in an indirect way, the velocity distribution of hydrogenic ions
along the sight line.
Neutral particle analyzers are in general used to measure the ion temperature and
distribution function in the plasma center, particularly during ICRF or neutral beam
heating[7].
When the NPA was installed on Alcator C-Mod, with the intention of monitoring
the central ion heating, it was found that the signals changed much too fast during
ICRF heating to represent central heating, and that the apparent temperature implied
by the signals was not consistent with results from other diagnostics.
We realized that this data represented an interaction between the RF heating
wave and the plasma edge, and that it was important to determine all the effects
the edge heating may have on the plasma and to understand what mechanism was
generating the fast ions.
1.3
Edge Ion Heating
The edge ion heating observed on Alcator C-Mod is characterized by a large increase
in the flux of neutral particles with energies above a few keV. The increase occurs
when the ICRF power is applied, and is much faster than other changes in the plasma.
The edge heating has a threshold in RF electric field, which is sensitive to changes in
temperature, density, and magnetic field in the plasma edge. The data suggest that
the edge heating occurs in the near field of the RF antenna in the outer scrape-off
layer of the plasma.
Ion edge heating in Alcator does not generate impurities, and does not affect the
coupling of RF heating power to the plasma. The total power absorbed by the edge
through this mechanism is at most a few percent.
1.4
Parametric Decay Instability
RF probe spectra typical of parametric decay instability (PDI) [8] have been observed
during RF heating on C-Mod. The RF electric field threshold for decay into an ion
cyclotron quasimode and an ion Bernstein wave have been calculated for C-Mod edge
plasmas. The change in observed edge ion heating at different parameters is consistent
with predictions of PDI theory, particularly the dependence on the magnetic field in
the edge.
1.5
Outline of this Work
Throughout this thesis, two threads are, by necessity, intertwined. The first is the
study of the effects edge ion heating has on the plasma, and the second is the search
for a mechanism responsible for edge heating. Because certain topics, such as particle
dynamics, are relevant to several parts in this work, topics are grouped thematically:
Alcator C-Mod and its diagnostics; particle dynamics and confinement; observations;
causes of edge heating.
More specifically, Chapter 2 of this thesis describes the basic operation of Alcator
C-Mod including the details of diagnostics and systems that are specifically referred to
later. The data from the neutral particle analyzer, covered in Chapter 3, is the basis of
this work. Chapter 4 presents dynamics of fast particles in the plasma edge, focusing
on energy and pitch-angle ranges of interest with respect to the energetic edge ions.
In Chapter 5, we review observations of fast ions in the plasma edge during ICRF
heating that have been made on other tokamaks. The various measurements of edge
ion heating and its effects that have been made on Alcator C-Mod are described in
Chapter 6. InChapter 7, the various mechanisms that can produce suprathermal ions
in the edge are considered. A brief introduction to parametric decay, the methods and
results of calculations, and their correspondence to the data are found in Chapter 8.
In Chapter 9, we summarize the work and draw conclusions.
Units and Conventions
All equations are expressed in the MKS system. Temperatures are assumed to include
the Boltzmann constant, so T in an equation is a value expressed in joules. In text
only, plasma temperatures are given in electronvolts (eV), as is customary in plasma
physics.
Unit vectors are represented by a "hat". For example 2 is the unit vector in the
x direction.
Chapter 2
Plasma operation on Alcator
C-Mod
2.1
Sample Alcator C-Mod Shot
The magnetic geometry and basic operation of the tokamak are well known[9]. Data
from a fiducial, shot 9602070041, is included as representative of Alcator data to
show typical time evolution of a shot and typical radial profiles. Discharges with
a particular set of controlled plasma parameters are frequently repeated to monitor
machine conditions; these shots are called fiducials. Fiducials use ICRF heating in
the hydrogen minority regime at Bo = 5.3 T. Figure 2.1 shows the time history of
several important plasma parameters during the plasma shot. Time zero is when the
toroidal electric field that drives the plasma current starts; the plasma begins to form
within a few milliseconds after. From 0.5 to 1.1 s, the plasma is completely formed
and parameters are stable; this period of time is the "flat-top". The ICRF heating
is turned on at 0.79 s, and central ion temperature starts to rise. At approximately
0.84 s, the plasma enters a regime of improved energy and particle confinement, called
1The
number of an Alcator discharge shows the date of the run, and which shot that day.
960207004 is the fourth shot on Feb. 7, 1996.
Shot 960207004
.
BTin
.............
..............
6 : .....................
T
Central Te (ECE) inrkeV
2.0 . .
,A
--.
.............
i.........
:.
..........
......-- 4:.-............. .............. ......
0
io
3.o
0.20
0.50
1:0
a
Plasma current
in:MA
1 . ........
.......
__
0:5~
..............
L0:
----------
---
--------
..
2:0 -..
5-------------------------------0.0
.o
--------------1;5
0.0O
0.50
1:0
1:5
SCentral
Ti(neutrons)
In
keV
.
..........
...
,2.0 .--:............. •.............
-----.....
2.-
.
....0..........................
o.po
o.o
1o0
1;5
m- 3
Central electron density In 1020
--- - - - - --- ----------------
4 : --
0.§o
lio
15
: Main plasma
bolometEr in .........
MW
..............
:6 .......
1-5
3:---------
--------5-
2:
------------1
..
.-------------------------------0.00
0. O
1'.0
1i
-3
Average elertron density:in 1020 m
I
i
I
4 : -- --- - - - - --- - - - -
2
o .o
.............---... ..
0:2 .............
.0060.0-10
0.- ............. ............. ............. 1
O.b)O
I
-- r -
•
10:
--- ---------- --
0.50
i
1.-0
1.5
-i-
-i--
Global Do, emission in A.U.
............. ............. .............
.............................
- -
So
000
. .
. .... .....
0" ,
: ............
f.............
..............
%
1
1
.................................
030n
1:o
o.bo
o. o
1,0
1'5
Net RF power in MW
log(F-port
neutral
pressure
in
Torr)
.
.............
:..............•..............
....
-3
. .......... ............. .
.c
......-....
-------------------2.-i
i
,
3.-- .............
-----........
........-.
-. ....
:o.o
0:0O00
;
0350
1-0
Time in seconds
1R
-6. ........ --
Nohn
.............................
n .n
1n
Time in seconds
Figure 2.1: Time series data from fiducial, shot 960207004.
1-r
"H-mode". The confinement regime observed before 0.84 s is referred to as "L-mode".
Because of the improved confinement, the plasma density and temperature increase.
The level of impurities in the plasma also increases, as is shown by the increase in
power emission detected by the bolometer array. The plasma current and the B 0 are
ramped down starting at 1.0 s and 1.1 s, respectively.
Figure 2.2 shows midplane profiles of a few of the plasma parameters. The two
vertical dotted lines show the location of the plasma center, at R = 0.682 m, and the
last closed flux surface (LCFS), at R = 0.890 m. It can be seen that the current and
electron temperature are sharply peaked, while the density is fairly flat. The plasma
edge is much colder than the center, but the density at the edge is within a factor of
two of that at the center. Edge parameters change with a much smaller scale length.
Typical electron density and temperature outside the LCFS are shown in Fig. 2.3,
taken from fast scanning probe data[10].
The flux surfaces of this plasma are plotted in Fig. 2.4.
The plasma is fully
diverted, with a single lower null. The poloidal limiters are represented as the righthand side of the vessel on the drawing. These limiters are between F and G ports, and
between A and B. The flux surfaces are reconstructed by the EFIT code[11] using
data from magnetic pickup coils outside the plasma.
The last closed flux surface
(LCFS) is the outermost flux surface that does not intersect a material surface. The
region outside this, excluding the divertor, is the scrape off layer (SOL).
Figure 2.5 shows a top view of the machine showing the limiters and port lettering
system. Figure 2.6 shows the relative position of the RF antennas, the PCX, and the
limiters.
2.5*1020
Electron density (m-3 ) vs. R(m)
I1 I
I
T
1
'
I
1.5*107
Current density
, .
I
2.091020
.5*1020
1.001020
5.0.106
5.0.10
19
OL
0.4
Electron temperature (keV) vs. R(m)
' I ' " I'
0.8
Poloidal field (T) vs. R(m)
1.51
0.0 L
0.4
0.6
0.4
0.6
0.8
Figure 2.2: Midplane profile data from a fiducial, shot 960207004. Dashed line represents position of the magnetic axis; dotted line shows position of last closed flux
surface.
Edge density vs. p
1.4
1.2
E 1.0
0.8
0
*-
0.6
0.4
0.2
0.0
10
20
30
p in mm
50
40
30
20
10
0
10
20
p in mm
Figure 2.3: Midplane electron density and temperature profiles outside the LCFS.
p = 0 corresponds to the LCFS. Dotted line shows location of the limiter. Profiles
are fits to data taken by the fast scanning probe.
0.4
0.2
E 0.0
N
-0.2-
-0.4-
-0.6
0.4
0
0.6
0.8
1.0
R in meters
1.2
1.4
Figure 2.4: Flux surfaces from shot 960207004
2.2
Radio-frequency Heating
On Alcator C-Mod, up to 3.5 MW of RF power in the fast wave at 80 MHz is routinely
used as auxiliary plasma heating[4]. This is in addition to the ohmic heating from
the plasma current, which is in general on the order of 1 MW.
2.2.1
Heating Schemes
Both electron and ion heating schemes have been used on Alcator C-Mod[4].
During minority ion heating, the fundamental cyclotron resonance of the minority
species is placed in the center of the plasma. In deuterium majority plasmas, minority
heating of H at 5.3 T and 3 He at 8 T has been performed. During H-minority heating
at 5.3 T with the cyclotron resonance at the plasma center, greater than 80% of the
RF power is absorbed by the plasma. Ion heating can also occur when the minority
Figure 2.5: Top view of Alcator C-Mod cell, showing positions of RF system, neutral
particle analyzer, and other diagnostics
21
Figure 2.6: Top view of Alcator C-Mod, showing antennas, limiters, PCX sight line,
and showing port lettering system
22
Figure 2.7: Double-strap RF antenna: Left figure shows Faraday shield. Right figure
is a top view, showing current strap, Faraday screen, and protection tiles.
second harmonic is in the plasma, for example at 2.6 T with a hydrogen minority.
2.2.2
Antenna
There are two RF antennas on C-Mod, one in D-port and one in E-port. Drawings
of the antenna in E port are shown in Fig. 2.7.
Their centers are 0.57 m apart
toroidally. Each antenna consists of two current straps, grounded at the center, fed
out of phase at the ends. The straps are 0.102 m wide, and have a vertical extent of
0.416 m. The separation of the straps, 0.284 m, sets the dominant kl of the spectrum
of power launched. The straps are curved to closely conform to the outer edge of the
plasma. The two antennas are driven at frequencies that differ by 0.5 MHz in order
to eliminate interaction between them.
Both antennas have Faraday shields, consisting of rods 9.5 mm in diameter, offset
from horizontal by 100 to be approximately parallel to the magnetic field. The rods
are separated by 3.4 mm, and are covered with a non-conductive coating. The rods
are 7.9 mm in front of the current strap. The purpose of the Faraday screen is to
shield components of the RF electric field parallel to the total magnetic field, and to
protect the surface of the current strap from photons and energetic atoms and ions
from the plasma.
There are antenna protection tiles on the sides of the antennas that project 5 mm
beyond the rods in the Faraday screen.
These tiles in turn are 5 mm behind the RF protection limiters, poloidal limiters
located between G and H ports and between A and B port on the outboard side. In
general, the gap between the LCFS and the limiters is from 5 to 15 mm.
Combining these measurements, the surface of the Faraday screen is approximately
20 mm from the LCFS, and the surface of the current strap is approximately 40 mm
from the LCFS.
Directional couplers measure the power delivered to the antennas, and the power
reflected. The difference between these is the total power radiated by the antenna.
The fields in the coax that delivers the RF power from the transmitters to the antennas are measured by probes. Using these two quantities it is possible to calculate
the current in the current straps and the loading resistance of the antenna. These
measurements are made throughout every shot with a time resolution of 0.10 mins.
2.2.3
RF Fields in the Plasma Edge
The purpose of the antenna is to couple RF power into the center of the plasma, where
it propagates as a fast wave[12] toward a cyclotron resonance or mode conversion
layer. RF waves at 80 MHz have vacuum wavelengths too long to propagate inside
the Alcator C-Mod vacuum vessel.
In studying nonlinear effects, it is necessary to know the magnitude of the electric
field in various regions. There are no measurements, but the wave magnitude can be
estimated using known quantities: the RF power flux and the fast wave dispersion
relation.
Details are in Appendix A. At 500 kW of injected RF power, we find
that the electric field in the propagating part of the heating wave is approximately
12 kV/m in the plasma.
The fields from the antenna near fields are much larger than the fields due to the
propagating part of the wave, however. The near field is due to the self-inductance
of the antenna, and is not directly related to the power coupled to the plasma. Diagnostics in the antenna coax measure the maximum voltage in the transmission line,
which is approximately equal to the voltage at the top of the antenna. For a 500 kW
shot, the voltage from the top of the antenna to the center is around 25 kV. If the
voltage in the current strap were uniform, then the vertical electric field would be
approximately 120 kV/m. This is the best estimate that can be obtained without
extensive modeling of the antenna, Faraday shield, and plasma. Since the field falls
off on scale lengths much shorter than the antenna height, the near field of the current
strap is similar to that of an infinite bar. The field should be approximately constant
from the surface of the strap to 50 mm away, and then fall off like In H/d where d is
the distance from the current strap and H is the antenna height.
The Faraday screen causes even stronger fields in its vicinity[13][14]. The vertical
electric field is shorted out inside the rod, and intensified between the rods. Since
the rods are 9.5 mm thick, separated by 3.4 mm, the field between the rods is larger
than the average field by a factor of (9.5 + 3.4)/3.4 = 3.8. An average vertical field of
120 kV/m leads to a field of 456 kV/m between the rods. This falls off to the average
field in a distance on the order of the gap between the rods, 3.4 mm.
The approximate midplane profile of the electric near field is shown in Fig. 2.8.
The details of this calculation are in Appendix C. As discussed there, the screening
of the electric field by the plasma can be neglected in calculating the poloidal RF
electric field in the SOL.
At 500 kW the electric near field of the current strap is about ten times larger
than the field in the propagating wave, and the field near the Faraday screen is four
times larger than that. This means that various interactions between particles in
the edge and the RF power will be much stronger in the vicinity of the antennas.
400
I
300:5
E
Uj)
LL
200
"E
-J
0
ai)
U-
Ir
100
0-
0.88
0.92
0.90
Major radius in meters
Figure 2.8: Midplane profile of RF electric field near the antenna.
0.94
However, if the antenna is putting power into global surface modes or propagating
waves other than the fast wave, fields far from the antenna could be larger than
estimated above[15].
2.3
2.3.1
Diagnostics
General diagnostics
As a world-class tokamak experiment in regular operation, Alcator C-Mod has an extensive array of diagnostics[16]. Many of these diagnostics are a necessary part of the
control systems that operate the tokamak and shape and control the plasma. They
also allow comparison of plasmas to monitor reproducibility. These core diagnostics include an interferometer for measuring electron density profiles[17], a Michelson
interferometer measuring electron cyclotron emission to determine electron temperature profiles, and magnetic pickup coils to find the plasma position and shape. Ion
temperature is calculated from neutron emission by the plasma, and spectroscopically.
2.3.2
RF Probes
RF probes are used to measure the propagation of the ICRF heating wave and observe
a variety of nonlinear phenomena that can occur.
There are RF probes throughout the machine. There are two types. The "B-dot"
probes consist of a pair of loops. The signal from these is V = -NA
•fi, where
N is the number of loops of area A, and fi is normal to the area. The probes are
shielded from the plasma and wired like humbuckers to further prevent pickup of the
electric component of waves. There are also Langmuir probes that are effective at
radio frequencies to measure the electric component of RF waves. They are unbiased,
and the voltage on them is measured directly. The tips of these probes are tungsten,
and extend 2.5 mm beyond the molybdenum housing.
There are fixed Langmuir probes on the A-B limiter and the G-H limiter, above
and below the midplane. There are fixed loop probes on the antennas behind the
current straps. There are three sets of movable probes. In J-port, there are two
probe heads with a total of 3 Langmuir probes, 2 B-dot probes normal to the poloidal
direction, and 2 B-dot probes normal to the toroidal direction. These probes are
0.11 meters below the midplane and move horizontally. They are positioned within
approximately 10 mm of the limiter radius. There is a probe head with 4 Langmuir
probes mounted on the A-top port. It is at a major radius of 0.735 m. In A-port
there is a probe with 4 Langmuir probes. It is situated 0.10 m above the midplane,
and scans horizontally. Additionally, this probe can be scanned in and out during a
shot, measuring a profile. The scan takes about 20 ms.
Signals from these probes are digitized by fast digitizers or a spectrum analyzer.
The fast digitizers can be operated at a sampling rate up to 1 GHz. The spectrum
analyzer can be run in several modes. It can be used to take low resolution frequency
spectra continuously, it can take one higher resolution spectrum per shot, or it can
follow the time dependence of power at one frequency. In the higher resolution spectra,
the resolution is usually 300 kHz FWHM for a scan from 1 to 100 MHz.
2.3.3
Impurity Measurements
An important question addressed in Chapt. 5 of this work is whether the fast ions
produced in the edge during ICRF heating generate impurities. While there is a wide
range of measurements that are related to impurities, the bolometer arrays, the moly
monitor, or the Z-meter would show the results of impurities generated by the edge
ion heating, and are briefly described below.
Moly Monitor
The surfaces inside the vacuum vessel exposed to the plasma, including the divertor
plates, limiters, and the inner wall, are covered by molybdenum tiles. Alcator C-Mod
is unique in its high-Z metal plasma-facing surfaces. Because high-Z materials can
potentially radiate large amounts of power, the level of molybdenum in the plasma is
monitored. The diagnostic known as the "moly monitor" measures XUV light emitted
in spectral lines of high charge states of molybdenum, from Mo xv to Mo xxxii[18].
There are three multilayer mirror-based polychromator channels, measuring three
spectral lines at a time resolution of 0.2 to 1 ms.
Bolometers
Alcator C-Mod has arrays of bolometers viewing the main plasma and the divertor region[19]. For the main plasma, there are 16 channels which span the plasma
toroidally at the midplane. They are sensitive to neutrals and photons from visible
through x-ray energies. There are 8 bolometer channels viewing the divertor. The
measurements are inverted to calculate emissivity profiles, which are then integrated
to find the total radiated power in the divertor and main plasma.
There is also a diagnostic usually referred to as the "27r bolometer". This is in
fact a single, uncollimated XUV photodiode, and not a bolometer, as it is sensitive
only to photons. It gives a fast measure of the global radiation from the plasma.
Z-meter
The "Z-meter" is an array of detectors measuring visible bremsstrahlung emission
from the plasma with a fast time response, 10 ps. Using the density and temperature
profiles, the quantity Zeiff =- Ej niZ/ne, a measurement of the impurities in the
plasma, is calculated.
2.3.4
Neutral Density
The "Ratiomatic" pressure gauge, a standard Bayard-Alpert gauge[20], measures the
neutral density at the midplane far from the plasma. The Ratiomatic gauge has a
time response of 17 mins.
There is evidence that a cold, tenuous plasma fills the volume around the confined
plasma[10]. At typical parameters, Te = 10 eV and n, =1 x 1019 m - 3 , the mean free
path of room-temperature neutrals is 0.03 m, much less than the distance between
the vessel wall and the SOL. Because of this, the neutral pressure at the Ratiomatic
will not be the same as that at the last closed flux surface.
Inside the bulk plasma, the profile can be modeled with the computer code
FRANTIC[21], and spectroscopic measurements made over a sequence of identical
discharges have shown good agreement[22]. In these discharges, the neutral density
as measured by the Ratiomatic was 2 x 1017 m - 3 . The measured neutral density at
the LCFS was 1 x 1014 m - 3 , showing that there is considerable screening of neutrals.
We will consider the Ratiomatic measurement to be an upper limit on the neutral
density at the LCFS.
Chapter 3
Perpendicular Charge Exchange
Analyzer
Charge exchange reactions occur when an atom or ion captures an electron from a
neighboring ion or atom. For example, a hydrogen ion may capture the electron from
a nearby hydrogen atom. After this ion is neutralized, it is no longer confined by the
magnetic fields in the plasma, and it will leave the plasma in a straight trajectory
(barring collisions and reionization). Thus the measurements of the flux of energetic
neutral particles from the plasma provides information about the population of ions
in the plasma[7]. The particle measurements used in this study were made using a
neutral particle analyzer we refer to as the perpendicular charge exchange analyzer
(PCX). The PCX simultaneously measures hydrogen and deuterium flux in 39 energy
channels and can be scanned poloidally and toroidally.
3.1
Source of CX Neutrals
To interpret the raw data from the PCX, an expression must be developed that relates
the count rates of the different energy channels to the plasma parameters. In this
section, an equation is presented for the energetic neutral flux that arrives at the
PCX beam line in terms of the volumetric source rate and the attenuation of the flux
passing through the plasma.
In a charge exchange reaction, an ion captures an electron from another atom or
ion. If the energy of the electron doesn't change, then the reaction is termed resonant charge exchange, and the momenta of the nuclei are, to a good approximation,
unchanged.
Alcator C-Mod is operated with majority species of D, H, or 4 He, and minority
species of D, H, or
3 He.
The bulk of the experiments described here were done
with a D majority, H minority plasma.
Because the He gas in the stripping cell
is only efficient at stripping D and H the PCX on Alcator C-Mod has so far only
been used to measure these two species. Equations and discussion in this chapter are
therefore simplified to apply only to charge exchange reactions involving hydrogen
and deuterium ions and atoms.
3.1.1
Source Rate
The source rate of energetic neutrals with a given velocity, ',, is
S(i,7) -
d3vofi(N,)fo(o,)cA(|
-
i~o l
.
(3.1)
The distribution function for the ion species, D or H, is subscripted with i, and
the 0 subscript is for the total of the hydrogenic neutral species.
Since the PCX
measurements are made at energies much higher than the neutral temperature, the
velocities of the neutrals are not significant, simplifying Eq. 3.1 to
S(G,
) = fi (,
i)no(Y)a
(vi)vz.
(3.2)
The cross-section for the resonant charge exchange reaction for hydrogen is on the
order of 10-19 m 2 for the energies of interest. The cross section as a function of energy
Exchange Cross Section
10-18
E 10
0
10-2
10-21
101
103
102
104
105
Ion energy in eV per nucleon
Figure 3.1: Charge Exchange Cross-section for Hydrogen
is plotted in Fig. 3.1[23].
3.1.2
Attenuation
Energetic neutrals produced at some position Y in the plasma must travel through the
plasma to be detected by the PCX. As these are neutrals, they are unaffected by the
electric and magnetic fields. The neutrals may, however, be involved in another charge
exchange reaction or be ionized in a collision, and thus "lost" from the outgoing flux.
Collisions between neutrals can be neglected. For each point along the trajectory of
the escaping neutrals, we can calculate the attenuation, a, as a function of position
and velocity. If the flux is represented by F, then
dF =-F
dx
.
(3.3)
The quantity a is the reciprocal of the mean free path, and is a function of plasma
density and temperature. If the velocity of the escaping neutral, v, is significantly
greater than the local ion thermal velocity, then the contribution to a from chargeexchange is approximately noacx. Contributions from ionization reactions are more
complicated. If vte > v 0 , then the term in a is f d3 vei(ve)ve/vo. If x' is used to
parameterize the sight line, between the point I where the energetic neutrals are
produced and the point of observation Xobs, then the flux of neutrals reaching the
observation point is reduced to
S~b Sef~obs
f b ce(x')dx'
.
I')
Sobs-- 5 Se o s
(4
(3.4)
This is the source rate of CX neutrals from a unit volume. To calculate the actual
signal measured by the PCX, we must consider how the analyzer is constructed.
3.2
Apparatus
The heart of the PCX, the stripping cell and the main chamber, along with the
microchannel plates and much of the electronics, was used on TFTR and has only
minor modifications from the one developed and implemented on the PLT tokamak in
1984[24]. This instrument was brought to MIT, and new support structures, vacuum
systems, and control electronics were designed and constructed[25].
An engineering drawing of the PCX is included as Fig. 3.2. The parts of the PCX
are described below in the sequence that an incoming neutral particle passes through
them.
3.2.1
Beam Line
The beam line serves simply to connect the main part of the PCX to the Alcator CMod vacuum vessel. The beam line has four baffles, copper gaskets with holes 7/8"
in diameter, to reduce the influx of neutral thermal gas when the gate valve is opened
during a shot. With this hole size, the baffles do not reduce the energetic particle
flux to the detectors. The beam line has a separate vacuum pump. When neutral
pressure in the beam line is above 0.1 mTorr, significant scattering of the incoming
energetic neutral particles occurs and the data are not used.
or
0
CD
M
M
O
C.
0
Also, 13" tangential (toroidal) range
c..i
cJ
3.2.2
Stripping Cell
At the ends of the stripping cell there are two rectangular apertures that set the
etendue of the analyzer, limiting the incoming energetic neutrals to those with velocities very nearly parallel to the axis of the beam line. The collimation is necessary for
accurate energy and mass measurements in the analyzer. The apertures are 1.3 mm
by 2.4 mm and the length of the stripping cell is 0.248 m.
The stripping cell is kept at a constant pressure of approximately 1 mTorr of 'He
gas. The He gas serves to strip the electron from the incoming energetic neutral
hydrogen atoms. The efficiency of the stripping cell for hydrogenic species in the
region of interest is approximately 0.2, i.e. 20% of the incoming atoms are converted
to bare protons or deuterons. More generally, the stripping efficiency is a function of
incoming particle energy and the helium pressure in the stripping cell[26]. The He
pressure in the stripping cell is measured by a vacuum gauge during the shot, so that
the correct stripping efficiency is used in analysis.
3.2.3
Main Chamber
Particles passing through the stripping cell enter the main vacuum chamber of the
PCX. This chamber is made of iron to shield out other fields from the tokamak and
provide a return path for the magnet flux. Ion paths in the chamber are directed
by parallel electric and magnetic fields generated by a pair of biased plates and a
magnet. The fields are as close as possible to uniform, and have their axis oriented
in the toroidal direction. The magnetic field (B) causes the incoming ions to execute
one half a Larmor radius to the detectors, as shown in Fig. 3.3. This half orbit takes
time tj = 7rm/qB, regardless of particle velocity. The diameter of the half orbit is
r = vm/2qB, so ions with different velocities finish a half orbit with different vertical
displacements and strike different detectors.
At the same time, ions are accelerated parallel to the field axis by the electric field
0.5
.
I
I
Typical particle trajectory
/
Stripping cell
CO
....
...........
...........
-----0.0-----
E
C
N
Beamline to machine
-0.5 -
MCP detectors
PCX main chamber
- 1 .0
.
1.5
.
.
I
2.0
.
.
.
I
I ..
2.5
R in meters
.
.
I
3.0
.
.
.
3.5
Figure 3.3: drawing of PCX main chamber
(E). At the end of the half orbit, they are displaced by (qE/2mn)t2 = m7r2 E/qB2 .
Note that this is also independent of incident particle velocity. Therefore, hydrogen
ions are displaced along the magnetic field axis by a certain distance, and deuterium
ions are displaced by twice that amount, thus separating incoming particles by mass.
3.2.4
MCP's
Particle detection is by means of three arrays of microchannel plates (MCP's). Each
array has 3 x 25 detectors. These are arranged to give 75 energy channels and three
mass (or m/q) rows. The MCP's are biased so that there is a voltage difference of
2 kV between the two sides. An incident particle causes an avalanche of electrons
in the tiny tubes that make up the MCP, resulting in a measurable puff of charge.
Because of problems of crosstalk observed at Princeton, we are using only every other
channel, giving 13 useful energy channels for each mass on a MCP, for 39 energy
channels for each mass. The dead space between the active channels causes about
60% of the particles in the energy range not to be counted. With three mass rows, the
analyzer was designed to simultaneously measure hydrogen, deuterium, and tritium
flux. As Alcator C-Mod has no tritium, that row is left unconnected, except for one
channel used for noise detection.
The range in energies of the particles the PCX measures is controlled by the
magnetic field. Due to the geometry of the analyzer, particles reaching the highest
energy channel have 40 times the energy of particles reaching the lowest channel. For
example, the PCX might be examining flux in the range of 1 keV to 40 keV. The
lowest energy the PCX can reliably detect is around 500 eV; the stripping efficiency
falls sharply below 1 keV[26], making it impossible to observe particles much below
this energy. The upper energy limit is set by the maximum magnetic field the magnet
can provide and the significant decrease in charge-exchange rate coefficient above
50 keV. We have successfully observed hydrogen ions with an energy of 150 keV with
the PCX.
3.2.5
Electronics
There are transistor amplifiers close to the anodes of the MCP's. They are connected
to the MCP pins with only 5 cm of wire, and are in vacuum. In practice, these
buffer amplifiers are fragile, and some have failed. During a run, 10 or more of the
78 channels used might be inoperable due to failed buffers.
The signals from the buffer amplifiers travel through 2 m of triax cable to the
pulse amplifier and discriminator modules (PAD's). These transform the pulses into
clean triangular pulses of uniform height. The pulses are then passed to a bank of
scalers. These count the number of pulses in a time bin, from 0.2 ms to 2 ms in these
experiments, for 1024 time bins on each shot. The data are collected over a CAMAC
system and stored on disk after every shot. The MDSplus system is used to access
the data[27].
3.2.6
Movement
The PCX has significant superstructure that allows the analyzer to scan in position
while remaining pointed through the hole in the port. In the nominal position of
the PCX, the sight line is horizontal and passes through the plasma at the midplane,
perpendicular to the toroidal field. From this position, the analyzer can be scanned
either "toroidally" or "poloidally".
While scanning, the PCX pivots about a point
R = 2.02 m at the midplane. There are bars attaching the main chamber of the
analyzer to a gimbal on the port cover which keep the analyzer aligned during a scan.
In a toroidal scan, the PCX is moved on rails by a motor; the sight line remains
horizontal through the midplane. This allows the PCX to measure escaping neutrals
whose velocity vectors are at a finite angle with respect to the magnetic field. The
PCX can be scanned toroidally to 11.7' from perpendicular. Note that the angle the
sight line makes with respect to the toroidal field is not constant, but changes along
the sight line. This extreme sight line is tangent at a radius 0.41 m (R/Ro - 0.7).
In a poloidal scan, a motor-operated scissors lift raises the PCX. The sight line
remains perpendicular to the toroidal field, but now passes through the plasma below
the midplane. At the furthest extent, the analyzer can "see" just above the X-point.
While the sight line remains perpendicular to the toroidal field, it is not perpendicular
to the total magnetic field. With the PCX raised to its maximum height, the sight
line makes an angle of 11.9' with respect to horizontal. This sight line passes through
Ro of the plasma at 0.25 m below the midplane. While the PCX could be operated
in a position where it has moved both toroidally and poloidally, this is not done in
practice. The PCX only changes position between shots.
3.3
Absolute Calibration
The analyzer observes particles produced in a particular volume of the plasma, which
is essentially a very narrow pyramid. At the center of the plasma, R = .66 m, the
footprint of the sight line is 0.023 m vertically and 0.042 m toroidally. The area of
the footprint varies approximately as the square of the distance to the stripping cell.
The analyzer only observes particles whose velocity is such that they pass down the
beam line. If I is a unit vector pointing along the sight line, then the analyzer selects
particles with &O01. Practically, it admits particles within a range of velocities,
I '0xll| <
(0.0011 - 0.00052),
V0
where the range is due to the aperture being rectangular, not circular. From this
we find a solid angle of acceptance of Q = 6 x 10- 7 . The range is calculated at the
center of the plasma, and is proportional to the inverse of the square of the distance
to the aperture. This range of velocities is only valid at the center of the "pyramid"
volume of the sight line. At the edge of the footprint, the range of velocities accepted
goes to zero, so most of the contribution is from the center of the pyramid.
To
make the equations tractable, the reasonable assumption is made that variations in
plasma parameters are negligible over the footprint of the sight line, and the ion
distribution function is well represented by an average over the range in pitch angle.
For example, the scale length for variation in the bulk ion temperature in the z
direction is about 0.13 m, which is 10 times the vertical extent of the viewing area.
Experimentally, variations in distribution function with pitch angle have scales of
more than 50 - .09 radians, much larger than the resolution noted above, 0.0011
radians. The range in pitch angle of the particles accepted can therefore be ignored
as well. Because the analyzer is sensitive to a greater range of pitch angles near
the axis of the sight line volume, the effective footprint area A used is1 the total
4
area. Thus a volume and velocity-space integral of particles in the sight line can be
simplified;
d3v[...]
d3x
(3.5)
dl v2dvA[...].
If the distance from point Y to the analyzer is L, then A oc L 2 and Q oc 1/L 2 . The
quantity Af
is then independent of L (and v 0 ). This product is called the 6tendue.
It can be calculated very simply. If the apertures in the stripping cell are Ax by Ay,
and are separated by a distance D, then the 6tendue is Ax 2Ay 2 /D 2 . For the PCX,
the 6tendue is 1.5 x 10-10 m 2 .
Combining Eqs. 3.2 and 3.4 and integrating over the entire sight line through the
plasma, the flux into the PCX, as a function of velocity, is
F(vi) =
JdlA2v
fi(vi, 1,5)no(F)cx(vi)viefobs
a(x')dx'
(3.6)
It is a short step from Eq. 3.6 to an expression for the actual count rates of the
PCX channels. As discussed above, the stripping cell has a finite efficiency, c(v).
8
Also, the MCP's have a finite detection efficiency, EMCP(V), in the range 60-85%,
though this is not known to high accuracy. If a channel counts incoming particles
with velocities in a range v1 -v 2 , the number of counts per unit time is
c = jv2dvs(v)Mcp(v)(v).
(3.7)
01
Because F(v) varies slowly enough as a function of velocity, this can be written
C = Avc(v)CMCp(v)F(v).
(3.8)
Letting k refer to the various channels, each responsive to a range Avk centered on
vk,
the count rates for the channels are
Ck = AVkEsc(Vk)6MCP(Vk)A~Uca(Vk)Vk
J
Cx()dx'
dlfi(vk, 1,)no0F)efobs
(3.9)
On the Alcator C-Mod PCX, a typical count rate during ohmic plasmas is 3 x 10'
counts per second at a few keV.
If the ions are assumed Maxwellian, and profiles for ni, Ti, and no are known or
assumed, then the expected count rate can be easily calculated.
From Eq. 3.9, dividing out the known quantities outside the space integral, an
experimentally measured count rate can be used to calculate
°
fob
f(v) =
This function
f(v)
Ge
(X')dx'
dl fi(v,1, )no()efobs cx()d
(3.10)
is a useful starting point for further analysis.
There are two special cases of interest. First, if the plasma is small or low density,
then attenuation is negligible and the exponential term is approximately equal to
unity. Tokamak plasmas are hottest in the center, and at energies sufficiently above
Ti the spatial integral is dominated by the contribution from the highest temperature
along the sight line. If fi is a Maxwellian, then In
or Ti,mnax = -[d ln f/dS]-
1
= -S/Ti,max + c, with S -
"
2
Practically, we do this by fitting a line to an appropriate
range of In f and using its slope. This formula, used to calculate bulk temperatures,
is only an approximation, and more accurate ways of finding Ti from the CX flux are
well known[7].
This derivative can be calculated for any
f,
and gives us a quantity referred to as
the "tail temperature". This temperature is not related to average particle energy,
but to the relative number of particles at different energies.
The other special case of interest is a suprathermal population in the plasma edge.
If this population exists over some Ar at radius r, then
f(v) = no(r)fi(v, 1,r)Ar/(.
1).
(3.11)
If no(r) is known, then the PCX has measured the number of suprathermal particles
(in a certain energy range with a certain pitch angle) per unit area of plasma surface. Similarly, quantities such as the energy in the tail per unit surface area can be
computed.
3.4
Sources of Error
Three possible sources of error in the PCX measurements must be considered. First,
any counting measurement has a statistical variation, a type of random error. Second,
the level of pick up on the detectors must be monitored. Last, any systematic error
must be accounted for.
3.4.1
Counting Statistics
The number of counts in a period of time follows a Poisson distribution. In analysis,
the counts are generally summed over a sufficient time period to get a significant
number. The error is used to weight fits for Ttaii. The noise signal is also governed by
counting statistics, and has an error. So if a channel gets m counts in a time period,
and n of these are noise, the number of counts is adjusted to m - n, but the random
error is Vm, not
3.4.2
/m- n. This is significant when m - n and n are of similar size.
Pick-up
The PCX is susceptible to several sources of noise. The neutrons produced by nuclear
reactions in the plasma are detected by the MCP's, and hard x-rays from run-away
electrons are also counted. Also, pick-up from the RF system is possible. One channel
from the tritium row of the MCP's is hooked up, replacing the last D channel. As there
is no tritium in the plasma, counts on this channel indicate the level of noise counts.
This is subtracted from the other channels during data analysis. The signature of
pick-up is that it is uniform across all the channels. In general, except in the highest
energy channels of the minority species, the number of counts due to pick-up is
negligible compared to the actual particle counts. Pick-up from the 80 MHz ICRF
system has not been observed.
3.4.3
Systematic Error
One well known source of error in this type of analyzer is mass contamination between
the D and H rows. As the ratio of hydrogen density to deuterium density has been
measured to reach as low as 0.5%, a small amount of scattering of deuterium into
the hydrogen row can overwhelm the hydrogen signal. Similarly, a high energy tail
in hydrogen produces high count rates in the higher energy channels, causing counts
to leak over and contaminate the D row. There is a specific signature of this effect;
during significant leaking, plots of In fD and In fH versus channel number (rather
than energy), have the same shape. Because the H row looks at energies twice what
the D row measures, the shapes of these two flux curves are usually different. So if
Trail for hydrogen is almost exactly two times the Ttrail calculated for deuterium, mass
contamination is indicated. In practice, mass contamination is only observed at the
lowest hydrogen concentrations, which seldom occur.
There is systematic variation between different channels, which can be seen at
times when the count rate is very large, implying that the random error is small.
This can be due to differences in the MCP channels or buffer boards. We assume
that this systematic variation in a given channel is random across the channels, and
the fits are still valid.
Note that there is no way to do an in situ calibration of the 6tendue or other
efficiencies. We have roughly checked the alignment using a light source, and we
believe the calculated 6tendue is approximately correct.
3.5
Sample Data
As an example of the data regularly taken by the PCX, Fig. 3.4 shows data from shot
960112030. It includes raw data from the deuterium row of the PCX, and the general
plasma parameters from analysis of data from other diagnostics. Figure 3.5 shows
the data from the hydrogen row. The channels on the D row collect neutrals with an
energy from 0.8 keV to 14 keV, and the H row covers energies from 1.5 to 29 keV. The
counts shown are acquired in 2 ms time bins. This shot is at relatively low density
for C-Mod, with a central electron density of 1 x 1020 m - 3 . This is a very high
plasma density compared to other tokamaks, however. The trace "nl_04" is the line
integrated electron density through a vertical chord through the plasma center. Note
that, even at this density, the mean free path for an energetic neutral with 10 keV
is 0.1 m, so the path from the center of the plasma to the edge is more than two
mean free paths so attenuation is significant. The signal in the PCX D row labeled
"noise monitor" is the noise monitor discussed above. Throughout the discharge, the
rate of counts in this channel is less than 10 counts/s, while the channels counting
particles are receiving at least 50 counts/s. On the H row, though, the channels above
channel 34 are at the noise level throughout most of the shot. On this shot, the PCX
sight line is horizontal through the midplane, and scanned toroidally to 10.40 from
perpendicular.
The spike in the PCX signals at 200 ms is when the plasma becomes diverted.
At 0.7 s, the RF heating is turned on for 70 ms. The plasma temperature increases,
as does the total neutral flux. There is a sharp drop in the H, light at 790 ms,
PCX deuterium data 960112030
'
ChfanneL4
eoe:
---
. .'..
10
.
.
:-----
40--6-:
-
hannel 2 ......
.1-:
......
:iO-
10
Q..n..
-
l :----------5
----
16-
0.:
-------
150
0.o
' --..... -Qan
-
77
.
S nl o4 in o0" r1
"
... .......0+
0o
Do
,
0
0
ChnneLl.9-8.....--,- 10...- .-CianneL3l
-------.
0o
-
...
I.............
.
.
.
.
....
.
.
.
.
.
,...... .,!....
. e ....
,---
li1o i:
Plasma current ihMA
1
-..., 4
--
1
.......
so9.
:..,..;-10 ............
0eChn 0- 6id---10
0ianneL1a Chan
.. - . ..
s
150.
'1
. . --..
--..
. .-
......
-
L3
-
0.0
150
-
o
1.
...
0 .....
-
1
nL29;
.I ......
-- , !:
--..-0---10-.....---10 ---------.
-'.11--- .-
'5. -.. '- 209:-..
-
-r
"
..
----§
00
09ce.---
-209:
1......
.... - r-...neL ...0...-----.,.....l.
;......
Net RE power in.MW
-
*2-
.
I
.
.
.
.
i
I
.
309:
..... -C nneL 15 ----C anneL28 ......
3o-: --209 ....... :.
.--20-:
-------...
....-......
.01o-.... .......
..
. 10--
.
00.10
'I'
150
0 0
10 I'
QhanneJ 4. ...... ,209:...... C minneL.
--"--'2
-----0'd0
.
di09: v:..
-----.
.•
' . .. ,,,"
1,,r
----.
anne
"':( ..
..
"" 3"" ....
o:..
40.
-
;. 40.-:- - -. ChanneL27 -.. ......
o'20
:.......
......
--
0o
p0
L9-----
-
-C
-
0 -------n T I'O
I5
. ....... ...... ....... ,.
0o
0
1;5
0Po
~~ ~ ~
~o
1Oe
.. .."• , . -------.-:.. . .. :........
,
,
,1
0-6
6 ....... .CW"an L 13 ...... 1; ........
...... 00
Qh eLI.64
1.5e:
- 16e-'
CtilaweL36
...... ;
00.
10......
00..1....
0.
00
0
joe,Qtian~ 8---------Cat
L1--------------Cane3
500
.0
0- 6 -o -6§ - - - ' ' i
oe
- - 10
1 5
*.n- - - - - - - - - - - - - - - - - - 0..
...
. ..
..
. .nL21
. -. .
05---me0:..
Qbanniel -------Ch
-- . . .1
-.
.- .. ChanneL34 . .. . ,
000 0
5 0
00o010
00
1:0
155
16: .....-Qann
7L1.1.7
......
---- ChnneL24 ......
;-.lo--- - . -ChanneL33:......
,s0:
..
50-......--...-*5-----------------..-----. 5-
~
'+
...
:..
...
."......
o;
:
.......
:.
..".
..
0--
6
.
.6-
.
..16-
-
i:
.....
,.i,
....
...
....
.....
.....
-15'
1
6
1
10
0.
00.
be:- -61.
0S--
-
-0
11,59
...... (Zh r.'419 ...... ".1"6:...... C. el'
2...... .,l........ Ctk ..
I'M ...... "
Ch.. Li-i6
loe.
-4D
-.
-Chne2
. .
-20
0?0
-- V,---id
ic-
-
0--
6
50
d6
--1
Canet2
-
- -
--
sCao
I . .. -. . . . .
..
.
C
3hann
13.
"
.
-Chan-- L3
----
1.0
Mi
-)
r--------n
..
.
......
.
.
.
-
-.
-
0.5ha
o38,
37.
. . .
SD.britness in A U.
-
L 26 40-.- - - -- ChannM
--
n
-"
0-.0......
0-
In
Time in seconds
o hn
1.
Time in seconds
Figure 3.4: PCX D row and plasma parameters
1
...........
---
5- -
.
A n
Time in seconds
'
T,(neitrons)
in aeV
"..........
:......
.......
-
-
------20-L11
6:
------ ie:------
19----
in ke:...... .
S T,(CE)
:......
........
..
,...
n lzn
......
,.....
1'n
Time in seconds
1,C
PCX hydrogen data 960112030
40.-.------Channe
---
20 ........
20.
. ...
.
40.-:--------Chan
-20.-------Ghannel-------:
--
- -------- Ghan
...........------.
--- h:..........:..
..0. .. ----n 8 -1.: 0D: )a
2 -5---------
G
4.---aChan-------- .
0.--------an
----- -20.--40.-: -------- C an e ............
8:--------
h
16:
6.--------Gan
132:--------
0
2--------nl33
Ghannel-..........
han ..... :....
.........
0.........
. .-..----------.. . - ------.........
---...
-
------ -0.5--:-----;-------;-------- 2.20.-------;-- --.....
20.-
........
o.-
-
- ------ G an el
--.......-------...
;---.- - ---- *-
40.------han
20.
- ....__y
-
-
I--:-
-------
------- -
hannel-
.
..----.
. 2.o0-:
. ------ G annel9:
......
;
20.- :h
......
20 .......
h annell
-
:-
50 ....
T (T
Time in seconds
Ghan-
Cha
.-.......
-
0 .------0....:
~.r5
-
"'Te u'0:UU
------
-33:
---
39 .han
.......
35
1
36:
.o-D:u
" ;u
'
4-------------
--
-
----------
1
------
2
-----
--
.........
. Chan-
.
6........
----..
•
Ghannel-2: .........
------
id
.- -------
--------. .-------
----
-
13 .........
~-
00
22-
15
- .......
- 42. .. --. . Ghan
. : .. 32
. .......
;
-T
0.----------------
.. han
4
8-:
------------
0"."
0.- ------- -- han
6
----
40.--------- hanne 7-:-------- -o.-:------Ghannel-2: .------20.-------.
.
.
--i
30.-------:
------.:o---annel-2
30.-ha
....
:......
. -.-.
. han l-21:
•.
.
10----a------n--
---------
.hannel4
$;
0.0:
6
2---oou
:
1
127 -
I0.-
.......... - _.'"
0- .-------. hannel-3: ..........
- :.. .
-5-. .n.
Ghannel-2 :.........
V-)V
1;V
. .......
1:)
Time in seconds
Figure 3.5: PCX H row
Ghan
uc-.
o: )V-------- V.)V
3 :.........
1;
Time in seconds
1
D and H flux, 960112030
34
+ deuterium
E] hydrogen
32 30
28
26-
E
mE]
24
_
mm
22-
E-
20
18
0
5
10
Particle energy in keV
15
Figure 3.6: PCX D and H flux, In f, as defined in Eq. 3.10
indicating the plasma has gone into H-mode. The improved confinement causes the
plasma density to increase sharply.
In Fig. 3.6, the calculated values for In fD and In fH are plotted, with the error
bars representing the random statistical error. The curve for hydrogen is the same
shape as the deuterium curve, and lower than it by 3.4. This indicates that the ratio
of hydrogen density to deuterium density is e- 3 4 = 0.03. Because the count rates are
relatively high, the statistical error is smaller than the systematic error for several of
the channels.
The deuterium flux during the ohmic L-mode, RF heated L-mode, and H-mode
phases is shown in Fig. 3.7. Fitting a straight line to the flux between 8 and 14 keV
gives ion temperatures for these times as 1.4, 1.9, and 1.7 keV. The ion temperature
measured by the neutron detectors is 1.4, 2.4, and 2.0 keV during these time periods.
Note that the electron density and temperature are also different during the H-mode
D flux at 0.5, 0.75, 0.8 s, 960112030
34
* 0.5 s, L-mode
A 0.75 s, 1.5 MW RF
o 0.8 s, H-mode
32
d~ga~~
30
28
26
AA m
Emm
A
24
10
01
22
2018
0
I
I
5
10
1
Particle energy in keV
Figure 3.7: PCX D flux during L-mode, ICRF, H-mode
period of this shot, as shown in Fig. 3.4. A similar plot for hydrogen is shown in
Fig. 3.8. During the RF heating, the fit to the hydrogen flux shows a temperature of
3.4 keV. During hydrogen minority heating with a low hydrogen concentration, it is
expected that the hydrogen distribution function have a tail hotter than the bulk ion
temperature[12].
H flux at 0.5, 0.75, 0.8 s, 960112030
I
I
o 0.5 s, L-mode
A 0.75 s, 1.5 MW RFE 0.8 s, H-mode
28
ifi
26
24
+
m[
[]
22
20
181
0
I
5
10
Particle energy in keV
i
i
I
I
Figure 3.8: PCX H flux during L-mode, ICRF, H-mode
I
Chapter 4
Dynamics of Edge Ions
To interpret the measurements made by the charge exchange analyzers, it is necessary
to understand the motion of energetic ions in the plasma edge.
In the complex geometry of the tokamak magnetic field, ions change their velocity
both parallel and perpendicular to the magnetic field. Most but not all ions remain
near a flux surface.
In general, interaction of ions with ICRF near a cyclotron harmonic generates ions
with large perpendicular velocity[12]. Orbit dynamics are used to calculate at what
location in its orbit an observed particle has a purely perpendicular velocity.
From orbits and collision mechanics, the confinement times of ions in the edge can
be estimated. Estimated confinement times are used in Chapt. 6 to calculate the RF
power absorbed directly by ions in the plasma edge.
4.1
Behavior of Fast Ions in Tokamaks
The magnetic field geometry controls the motion of ions in tokamaks. Edge ions of
interest are in banana orbits, an effect of poloidal fields, and possibly ripple-trapped
orbits, a result of toroidal field (TF) ripple.
To understand the motion of ions in a tokamak, we must first describe the relevant
parts of the magnetic field geometry, including the poloidal field and toroidal asymmetries in the toroidal field. Next, the orbits of banana-trapped and ripple-trapped
ions are presented.
4.1.1
Tokamak Magnetic Fields
The magnetic fields in a tokamak are generated by external magnetic coils, currents
flowing in the plasma, and eddy currents flowing through other structures.
The largest magnetic field in a tokamak is the toroidal field (TF), which is generated by a toroidal solenoid. In cylindrical coordinates (R,z,O), the toroidal field
(B q) is approximately BO = (BoRo)/R in the plasma region, where B 0 and Ro are
constants. Because the field is generated by a finite number of coils, the resulting field
has a weak q5 dependence. The field generated by the TF coils is better represented
by[28]
T
=
BoRo
(1 + 6(R, z) cos(No)) .
(4.1)
The number of coils is represented by N, and the quantity 6 is referred to as the
toroidal field ripple. In Alcator C-Mod, the ripple is less than 0.004 everywhere in
the main plasma. In the area where 0.7 < R < 0.95 meters and |zj < 0.3 meters, the
ripple is approximately[29]
6(R,z) = 0.0038e
R-0.89
0044
(4.2)
where R is in meters. Alcator C-Mod has 20 TF coils.
The toroidal field ripple has two effects important here. First, it makes the radius
of cyclotron resonance surfaces a function of q. For example, we will need to know
Re such that
Wrf -
B(R ).
Wci=
m
Re must be a function of q because of the TF ripple. Second, as will be discussed
more fully in the next section, the break in toroidal symmetry can lead to particles
being trapped near the minima in the ripple and lost from the plasma.
The toroidal component of the total magnetic field, B. q, also has a contribution
from poloidal currents in the plasma. This is largest in the center and goes to zero
at the plasma edge, and is independent of 4.
The poloidal components of the magnetic field come from plasma currents and
external vertical field coils[9]. By definition, the poloidal field is always tangent to
the flux surface. On Alcator C-Mod, the poloidal field is of the same order outside
the plasma as inside and is around 1/5 the magnitude of the TF at the edge.
4.1.2
Ion Orbits
While the equations describing the motion of ions in tokamak magnetic fields are well
known[30], a brief discussion is presented to facilitate their use in analysis of the data
presented later. The PCX measures ions at different pitch angles and at different
points in the plasma. Analysis of the ion orbits is necessary to understand where
in physical and phase space these particles may have been accelerated.
Also, the
confinement of these ions near the plasma edge depends on where their orbits take
them.
In the roughest approximation, an ion in a magnetic field will follow a helical
path along a field line. In the measurements of edge ion tails, ions with energies up
to 30 keV were observed in fields as low as 2 T. These ions have a Larmor radius,
PL = vi/wi, of approximately 0.01 meters. This is the same order as density and
temperature scale lengths in the plasma edge, but is small compared to scale lengths
of the magnetic fields.
The motion of ions in the curved magnetic fields of the tokamak is best described
through the use of invariants of the ion motion. If the time and length scales for
changes in the magnetic field are large compared to those of the orbits, there are
quantities that remain nearly constant during particle motion. These are the particle
energy,
=1- mv 2 ,
2
(4.3)
mvi_
1
P=
2B
(4.4)
the particle magnetic moment,
and the angular momentum,
PO = mRvo + qRA
,
whereB = V x A.
(4.5)
Solutions of these equations consist of passing orbits and two types of trapped orbits,
banana orbits and ripple-trapped orbits. Equation 4.5 represents a constant of motion
only when the magnetic field is toroidally symmetric; ripple-trapped orbits result from
asymmetries in the toroidal field, and angular momentum is not conserved.
Banana orbits
Consider a particle on the midplane at some R 1 > Ro, with a velocity Vl.
magnetic field B =-
&
,
In a
if p and E are constant over an orbit, then B and vj must
be related by
v
2
= 2 ( - Bp).
m
(4.6)
Writing in terms of the initial perpendicular velocity, total velocity, and particle
position (radius),
2
V=
1
v2
v 21
Ri
2
R
v
1
Rb
Rb
(4.7)
R
If the field line the particle follows reaches Rb, defined by
Rb
R
(
(
1
V
)
2
(4.8)
then at R = Rb the particle's parallel velocity goes to zero, and the particle bounces
back toward larger radius in a banana orbit. If the field line does not pass inside Rb,
the particle passes around the torus, making a poloidal rotation every few toroidal
rotations.
As a particle passes around an orbit, it deviates from a field line according to
Eq. 4.5. The width of the banana orbit is approximately
AR = 2mR v11
qB '
(4.9)
where Bp is the poloidal magnetic field. This equation is valid when AR is small compared to scale lengths for the poloidal field and radial electric fields can be neglected.
For thermal ions in the plasma edge, AR ~ 2 mm, depending on the plasma current;
this is calculated for a typical current, Ip = 800 kA implying Bp _ 0.6 T. Ions heated
in the edge plasma during ICRF injection have been observed with V1 = 0.5 at the
midplane. If such a particle has an energy of 10 keV, then the banana width of its orbit is 0.06 meters. This is large compared to relevant lengths in the plasma edge. The
distance between the LCFS and the RF protection limiter at the outboard midplane
is usually less than 20 mm. For these orbits, Eq. 4.9 is only a rough approximation.
The time for a particle to execute a banana orbit, Tb, can be given approximately
.'
.
"" ' . . . . . . .
..........
" . . . . . . . . . . . . . . . . ... . .. . . .
Figure 4.1: Banana orbit of an energetic particle
in terms of the basic parameters;
b-
2r 2rR B
vi
Bp
(4.10)
A banana orbit for an energetic particle in the outer plasma is shown in Fig. 4.1.
The figures on the left show two projections of the orbit in space. The figure on the
right shows the same orbit with the toroidal angle dependence suppressed.
The banana orbits of edge ions observed by the PCX are shown in Fig. 4.2. The
orbits are shown as viewed from the outboard midplane on a flat projection. This
figure also shows graphically that the PCX observes ions at their banana tips during
a poloidal scan, and at the center of their orbits during a toroidal scan.
Side View of Bounce Orbits
0.2
0.1
0.0
-0.1
-0.2
-1
0
Limiter
1
Antenna current straps
+ where analyzer sightlines intersect LCFS
banana orbits of particles viewed
magnetic field lines in projection
Figure 4.2: Orbits of ions in the edge
2
Ripple-trapped orbits
Banana orbits occur when vII = 0 along an ion trajectory, as described in Eq. 4.7,
at which point the particle reverses direction. The asymmetry in the more accurate
form for B 0 , Eq. 4.1, adds more possible bounce points, which lead to ripple-trapped
particles. Because the ripple in the toroidal field is small, ripple-trapped ions have
vi < v .
A particle can be ripple trapped if there exist two local maxima in the magnetic
field along the field line. Maxima occur when
VB -B=0.
Using Eq. 4.1, and noting that the dominant terms in VB come from the B dependence on R and 4,
__# BoRo
VB
R--R
D2 -
B -B
BoRo
R2
R
BoRo
R2 Nsin(NO)O,
R2
(4.11)
BoRo
2
R
(4.12)
sin(NO)BO.
Setting the left-hand side equal to zero shows the condition under which rippletrapping can occur,
BR < 5NB¢.
(4.13)
Defining 0 as the angle between Bp and 2, BR = B, sin 0, the condition becomes
sin 0 < 6N B
Bp.(
(4.14)
If the plasma has a circular cross section, then 0 is the poloidal angle. In the shaped
C-Mod plasmas, there is no simple interpretation to 0. In the volume where rippletrapping is possible, only a small percentage of the particles are ripple-trapped. Outside this region, no particles are ripple-trapped.
To find the maximum parallel velocity of a ripple-trapped particle, we use Eq. 4.4,
evaluated at the bounce points and the center of the ripple well,
my211max =
B(1 - 6),
-
(4.15)
2
(4.16)
0 = S - B(1 + 6).
Subtracting these and rearranging gives
V 11max
V
%
.
(4.17)
V
If 6 = 0.004, then particles with V"m--less than 0.009 can be trapped, that is with
velocity vectors within 50 of perpendicular to the magnetic field.
For ripple-trapped particles, the toroidal component of the vector potential, AO, is
not even approximately independent of 0. Because of this, the ion angular momentum
P0 is not conserved, and the particle is not kept near a flux surface, as banana-trapped
particles are. Ripple-trapped particles will instead exhibit VB-drift. In C-Mod, the
VB-drift direction is usually downwards. The drift velocity is
mv 1 B x VB
B2
2 qB
VVB
mvj
2
1
qRB(4.18)
Curvature drift is negligible for ripple-trapped particles, as they have vi > vil. For a
10 keV ripple-trapped particle, the drift velocity is 3 x 103 m/s.
Because the ripple-trapping region is small, particles can easily scatter out, with
a characteristic time smaller than the large-angle collision time by a factor of 6. At
the LCFS, thermal deuterons trapped in the ripple well will only drift a distance
on the order of 10-8 meters before scattering out. As will be discussed in the next
section, suprathermal ions have a much slower collision rate. Nevertheless, rippletrapped 10 keV ions will only drift approximately 1 millimeter before scattering out
of the ripple-trapping region of velocity space. This assumes that n = 1020
m
- 3
at
the LCFS.
4.2
Confinement of Edge Ions
The distribution function of energetic ions in the plasma edge is only partly a function
of the interaction with the RF fields and the orbit dynamics. The evolution of the
distribution function also depends on the rate at which these energetic ions either
escape from the plasma or lose their suprathermal energy.
There are three primary mechanisms by which the ions of interest are lost. They
may be unconfined, and not stay in the plasma for an entire orbit; they may charge
exchange and become energetic neutrals; or they may lose their energy through collisions.
In the plasma edge, the confinement times for these processes change on short
length scales. As we do not know exactly where in the edge the ion tails exist, we
must consider how the loss mechanisms will vary across the plasma edge.
4.2.1
Unconfined Orbits
A particle is unconfined if its orbit intersects a material surface. For the particles
viewed by the PCX the relevant material surfaces are either the outboard limiter or
the protection limiters on the sides of the antennas. As noted above, the banana
width and Larmor radius for particles in these highly energetic tails can be larger
than the width of the SOL, causing particles originating at even the LCFS to be
scraped off by the limiters.
For purposes of estimation, we will consider the lifetime of a particle on an unconfined orbit to be approximately one quarter of the bounce period, or
10 keV deuteron, this lifetime will be 23 ps. Recall that
4.2.2
b oc
I v/
Tb/4.
For a
(Eq. 4.10).
Collisions
There are three types of particle collisions to consider: collisions with ions, electrons,
and neutrals. At ion energies above a few hundred eV ion-neutral collisions will be
charge exchange reactions; these are covered in the next section.
The canonical form for collision frequencies is[31]
dv
dt = -e0"v,
(xa/l3)vuo I3
zi1' =(1+ m0
/
=
= m vv
(4.19)
/2T,
(x
>1)/(x-
= 47r
/ol
<
1,
~4 3 m?,3
1)
30
The condition X <K 1 holds for tail ions colliding on electrons, and x > 1 for tail
ions on thermal ions. Whether tail-ion or tail-electron collisions dominate depends
on the energy of the tail ion in question. Using k to refer to tail ions, and assuming
hydrogenic ions, the collision rates are
(+
Wk/i
mAcvk
mi4o
kj
)
2 3,4.0
4
mk
Mk/e
-
m
me
e4
e
3/2
m
A
3332e
2
mk/2T
(
)
(4.21)
Comparative plots of the different collision times are in Sec. 4.2.4.
4.2.3
Charge Exchange
The charge exchange reaction is described in Chapt. 2. Because the density of neutral hydrogen decreases rapidly in the SOL, the charge exchange rate has a strong
spatial dependence. Tail ions have large banana widths and large Larmor radii, and
will pass through regions of much higher neutral density then that at their guiding center/bounce tip. The actual neutral density is neither toroidally nor up-down
symmetric[22].
4.2.4
Discussion
We consider all the above processes to estimate the time that a suprathermal ion will
remain in a perpendicular tail. Ion-ion collisions do not remove ions from the plasma,
but they do scatter them out of the tail.
Figure 4.3 shows the collision frequencies for the various processes described above
in the SOL of shot 960829023. The x-axis is distance from the LCFS in mm at the
midplane. The temperature and density profiles are taken from data obtained by the
fast scanning probe, and are shown in Fig. 2.3. The neutral density is estimated from
measurements by the PCX during the ohmic heating phase of the discharge. Simulations using a Monte-Carlo code show the neutral density in the SOL is approximately
ten times that at the LCFS[32]. For this shot, the gap between the LCFS and the
outboard limiter is 12 mm.
The plots in Fig. 4.3 show that loss by hitting the limiter is the fastest loss mechanism for particles in orbits that intersect the outboard limiter. Depending on radius
and energy, this loss time is 3-100 times faster than for charge exchange, implying
that only 1-30% of the fast ions charge exchange before hitting a limiter, and only
this small percentage can be measured by the CX analyzer.
Deuteron with 0.50 keV
Vb
X
v C............
3
10
Vk e
.
-
2
10
20
10
)
31
p in mm
Deuteron with 1.00 keV
Vb
............................................................................
Vki
- - --
- -- - -
- - -
- -
-
- -
- -
- -
-
Vke
-
- - -
-,-.,-.
p in mm
Deuteron with 10.00 keV
05
00
04
--
--
vb
-
V .
03
2
ke
0
00
p in mm
Figure 4.3: Collision frequencies for deuterons at 0.5, 1, 10 keV, showing bounce
frequency (vb), charge exchange frequency (vex), and frequencies for collisions with
thermal ions (vki) and electrons (vke). X-axis shows distance from LCFS.
63
Because of the large Larmor radii and banana widths of the energetic edge ions
under discussion, we will assume that their orbits intersect a limiter. In this case, the
confinement time for energetic edge ions is on the order of tens of microseconds.
Chapter 5
Previous Observations of Edge Ion
Heating
It has long been known that the interaction between particles and waves can lead to
non-Maxwellian, non-isotropic velocity distribution functions[33]. The RF can add
energy to a small part of the ion population, creating a "tail" [12]. An ion tail in
the plasma center improves the damping of the fast wave, and, in a reactor, would
increase the reaction rate.
Fast ion populations in the center of an ICRF heated tokamak were first observed
on T-4[34]. A mass-resolving charge exchange diagnostic observed a large tail in the
hydrogen population. The analyzer could only measure the perpendicular distribution function. Variations with pitch angle of the ion distribution function were first
measured on PLT, using a scanning passive charge exchange analyzer[35].
Hydro-
gen tails in the main plasma during minority heating have been observed on many
machines[30].
These results are important in understanding power absorption during minority
ICRF heating in tokamaks. Indirect measurements of ion distribution functions near
the plasma center are made, though extensive modeling may be necessary to under-
stand the results[36]. Certainly there are unresolved questions regarding the transport
of the tail ions, including spatial diffusion of fast ions and response to transport events
such as sawteeth.
In order to heat the plasma center, RF power must be coupled through the plasma
edge. This coupling can be strongly affected by the edge conditions[6]. The strong RF
fields can also interact with the edge plasma ions through various linear and non-linear
mechanisms, and transfer energy to them. It is not surprising, then, that fast ions
in the plasma edge have also been observed during ICRF heating on tokamaks. The
mechanisms of edge ion heating and its possible effects will be discussed in Chapt. 7.
In particular, suprathermal ions have been reported on ASDEX, TEXTOR, JT60, and Alcator C-Mod during fast wave heating, and on DIII-D during ion Bernstein
wave heating.
During ASDEX experiments comparing optically-open and optically-closed Faraday screens, deuterons with large perpendicular velocities were observed when the
antenna with the optically-open screen was used[37]. The tokamak was operated at
2.2 Tesla. The RF power consisted of up to 1.2 MW second harmonic hydrogen minority heating at 67 MHz. The plasmas were run in double-null configuration, with
Ro = 1.67 m, a = 0.40 m, he
=
(2 - 6) x 1019 m - 3 . The deuterium flux showed fast
rise and decay times, 1-3 ms, and was very sensitive to toroidal field, leading them
to conclude that the flux originated in the plasma edge. The tail began at energies of
4 keV. The flux showed large fluctuations uncorrelated with any macroscopic plasma
parameter, including impurity production. The deuterium tail was measured using
a charge-exchange analyzer midway between the two antennas, perpendicular to the
magnetic field at the midplane.
Measurements were also made of fast particles using carbon implantation probes
in the center of an RF antenna on ASDEX[38]. Each probe was exposed during one
shot, and the depth distribution of particle implantation was measured using "elastic
recoil detection". Control probes, 1.5 m from the antennas, were also used. Particles
with incident energy up to 15 keV were observed. The deposition of suprathermal
ions was approximately twice as large for the probe in the antenna center compared
to the control. However, the deposition on the control probe was ten times larger
than the prediction based on measurements by the charge-exchange analyzer. From
this they concluded that most of the particles striking the carbon probes were ions,
not neutrals.
In the discussion of results from parametric decay instability investigations using
RF probes on ASDEX, they suggested that PDI could be generating the fast ions.
However, no evidence of correlation between edge heating observations and PDI as
observed by RF probes was presented[39].
Investigations of edge effects during ICRF heating on TEXTOR showed the presence of fast ions in the SOL[40]. TEXTOR was operated at a toroidal field of 2.02.6 T, with fi < 6 x 1019 m
- 3 ,R
0
= 1.75 m, a = 0.45 m, with limited, circular
plasmas. Laser induced fluorescence was used to measure Fe atoms sputtered from a
small, movable steel surface. From measurements at different plate positions they determined qualitatively that a population of fast ions existed in the SOL with 1.6 MW
of ICRH at 29 MHz. The sputtering yield increased quickly, in less than 1 ms,
compared to time scales for changes in the flux at the limiters or the bulk plasma
parameters, around 100 ms. The increases in SOL temperature measured with probes
were insufficient to account for the increased sputtering. RF sheaths were also ruled
out. Further investigations indicated that the ion tails were in the perpendicular direction. Evaluation of measurements of heat flux to the limiters during ICRH showed
changes in the heat flux too large to be accounted for by changes in n?,Te, and the
parallel ion sound speed, indicating an increase in the perpendicular energy of ions
in the SOL. No variations with plasma parameters or RF power were reported.
Fast ions were observed on the JT-60 tokamak with charge-exchange analyzers
during hydrogen minority heating[41].
JT-60 was operated with a divertor at the
outboard midplane, with BT = 3.7 - 4.3 T, Ro = 3 m, a = 0.93 m. These measurements were made with 1.5 MW of RF power at an unspecified frequency in the range
110-130 MHz. The charge-exchange analyzer used was below the midplane on the
outboard side with a sight line passing near the plasma center, and measured the flux
of ions with energies of 3.9 keV and 5.6 keV. The CX flux "jumped" with the injection
of ICRH, but only when the antennas were operated with (0,0) phasing. The jump
in flux did not occur when the antennas were operated in (0,wr) phasing. Parametric
decay into an ion cyclotron quasimode and an ion Bernstein wave was also observed
with RF probes on these shots. The toroidal field was scanned at constant RF power,
and the IBW amplitude and the jump in flux were found to be positively correlated,
leading them to conclude that the parametric decay generated the energetic ions.
Radiation losses from the plasma were also correlated with IBW amplitude, implying
that the PDI led to increased impurities in the plasma.
On the DIII-D tokamak, experiments with direct-launch IBW heating of the
plasma showed significant production of fast particles in the edge, as measured with
a toroidally-scanning charge exchange analyzer at the midplane[42].
experiments BT = 1.0 - 2.1 T and he = 0.7 - 5 x 1019
m
- 3
During these
. The IBW was launched
at 36 or 38 MHz by a loop antenna with an opaque Faraday shield at powers up to
1 MW. While this experiment did not involve fast-wave heating, the antenna structure is similar to that of a fast-wave antenna, the ratio wr,/WcD is similar to that
used in some of the Alcator C-Mod experiments, and their data is extensive. They
found that edge ion heating only occurred in inner wall limited discharges with low
densities (ie < 3 x 1019 m- 3 ), and not with higher densities, or in outer wall limited
or diverted shots. Scanning the CX analyzer toroidally, they found no parallel ion
heating. They observed perpendicular tails in both hydrogen and deuterium, with
tail temperatures of 5.5 and 1.8 keV respectively, significantly higher than the 0.8 keV
bulk plasma ion temperature. The tails formed only at RF powers above 240 kW.
The conclusion that these ions are generated in the edge was based on the fast rise
and fall times, less than 1 ms, and that the appearance of the tail had no effect on the
core ion temperature, plasma stored energy, or neutron production. During toroidal
field ramps, the sizes of the ion tails were shown to peak strongly when the resonances
w f = 5WcD and wrf = 4 WcD passed through the plasma edge.
These experiments measured parametric decay with an electrostatic RF probe
located 500 toroidally from the RF antenna. The probe was located at the midplane
at the same major radius as the RF antenna. The power thresholds for PDI and ion
edge tail productions with wrf =
4
WcD
near the edge were both found to be about
200 kW, and the tail amplitude was well correlated with the amplitude of the PDI
decay waves. The PDI and tail production were sensitive to relative amounts of
hydrogen and deuterium in the plasma.
While the focus of the present work is fast wave heating, it is worth noting that
ion tails have been observed in conjunction with parametric decay on tokamaks using
lower hybrid heating. On Alcator C[43], edge ion tails toroidally localized near the
launcher were observed during LH heating using a CX analyzer. Ion tails in the bulk
plasma caused by PDI were measured for the first time on the ATC tokamak[44].
The reported data regarding edge ion heating during ICRF heating on tokamaks
leaves several issues unresolved. There is no record of these observations on several
major experiments, including JET and TFTR. This may mean that no diagnostic
was used which would be sensitive to energetic ions in the edge, or that they were not
monitored. Also, the reported data are inconsistent regarding impurity production
during edge ion heating. The control of impurities is crucial to tokamak reactor
operation, and clarification of the role of edge ion tails is important.
The observations from different machines do have important similarities. Edge
tail production appears sensitive to the antenna geometry, the edge conditions, and
the magnetic field. Of course, there are several mechanisms by which edge ion heating
can occur (as described Chapt. 7), and the particular mechanism operating in any
one of these experiments is not necessarily that which is occurring in the others, or
in Alcator C-Mod.
Chapter 6
Observations of Edge Ion Heating
on Alcator C-Mod
6.1
Introduction
This chapter describes our observations of edge ion heating in Alcator C-Mod and an
analysis of its effects on the plasma. The rest of the present work will study possible
causes of the edge ion heating on Alcator C-Mod.
This chapter serves three purposes. First, it defines and describes the phenomena
of edge ion heating as observed on Alcator C-Mod. Second, it presents details which
will be important in later chapters as comparisons to possible edge-heating mechanisms are analyzed. Third, it details the effect edge ion heating has on the rest of the
plasma.
The perpendicular charge exchange analyzer was installed on Alcator primarily to
measure the central ion temperature, and more generally the distribution functions of
hydrogen and deuterium in the plasma interior. The latter are particularly important
in analyzing ICRF heating results[36].
The PCX takes data on essentially every
plasma shot. In certain cases, the PCX has measured ion temperatures in ohmically
heated plasmas, but in general the plasma is too dense, leading to high attenuation
(see Sec. 3.1.2), which reduces the energetic particle flux from the plasma center to
below-measurable levels.
During RF heating on Alcator C-Mod, the PCX often records a sudden large
increase in signal coincident with the RF pulse. The signal represents a large flux of
neutrals in the energy range the PCX, usually on the order of 1 to 20 keV. Because
the flux includes neutrals with energies well above the plasma temperature, we refer
to them as energetic neutrals. The energetic neutrals arise via charge exchange from
ions in the plasma. These ions are referred to as suprathermal, and make up an
ion tail, that is, a high energy component to an otherwise Maxwellian distribution
function.
There is a specific set of characteristics that define the jump in neutral flux: the
fast rise and decay times; the lack of correlation with macroscopic plasma parameters;
and the sensitivity to edge plasma conditions (Sec. 6.2). Analysis of these characteristics leads to the conclusion that the energetic ions are located in the plasma edge.
Specifically, some effect related to the RF heating generates highly energetic ions in
the plasma edge, and these ions charge-exchange and escape the plasma as energetic
neutrals. In the present work, "edge heating" refers to the generation of suprathermal
ions in the plasma edge during RF power injection.
The properties of edge ion heating depend on the RF heating regime. During hydrogen minority heating, there is a threshold electric field for edge heating (Sec. 6.3).
If the magnetic field is at a value such that the RF frequency is an exact cyclotron harmonic in the plasma edge, large edge heating at low RF power is observed (Sec. 6.4).
At the transition from L-mode to H-mode, the plasma edge changes, and this is
reflected in the energetic neutral flux (Sec. 6.8).
Theoretically, edge ion heating may affect the efficiency of RF heating. The edge
heating seen on C-Mod, however, has no effect on the coupling of RF power to the
plasma (Sec. 6.7) and it is not a significant channel of parasitic power loss (Sec. 6.9).
It also does not lead to an increase in the level of impurities in the plasma (Sec. 6.6).
6.2
Identification as Edge Heating
As discussed in Chapt. 3, the PCX is sensitive to neutrals generated at any position
along its sight line, and the flux measured can be written as a line integral, as in
Eq. 3.6. The radial location of the source of any particular particles must be inferred.
There are three characteristics of the particle flux that indicate that the energetic
ions are being generated in the plasma edge: (i) the fast time scale, (ii) the slope of
the flux, and (iii) the dependence on edge conditions.
6.2.1
Time Scale
On some discharges, the PCX records an increase in the flux of suprathermal neutrals
of up to a factor of 10' when the RF power turns on. The increase occurs on time
scales shorter than 0.2 ms (the PCX time resolution), much faster than changes in the
core ion and electron temperatures, which typically change on a time scale of 10 ms.
The decay time after the RF power turns off is similarly short. Again, in the center,
the temperatures decay over tens of milliseconds.
As described in Chapt. 4, and shown in Fig. 4.3, particles in the SOL with energy
in the keV range can have particle confinement times less than 0.1 ms. The short
decay time of the increased flux is typical of times in the plasma edge, rather than
the plasma center.
6.2.2
Tail Temperature
The agreement between spectroscopic and neutron-rate measurements of the plasma
central ion temperature shows that the distribution function of the majority ion
species in the plasma remains nearly an isotropic Maxwellian during ICRF heating
on Alcator C-Mod. Line widths in the atomic spectra of impurities in the plasma center measure the spread in energy of the distribution function. The impurity ions are
strongly coupled to the bulk ions via collisions, and have the same temperature. The
neutron production rate (for D-D reactions) increases with energy, so total neutron
production is in effect a measure of a higher order moment of the distribution function, and is therefore particularly sensitive to suprathermal tails. Ion temperature is
calculated from the neutron production using the assumption that the deuterium distribution function is a Maxwellian. The two measurements of central ion temperature
generally agree, implying the deuterium distribution function is a Maxwellian.
During the jump in flux, the value of Tail inferred from the neutral flux (as
defined in Sec. 3.3) shows no relation to the bulk ion temperature as measured by
other diagnostics. In shot 960227042, shown in Fig. 6.1,
the deuterium flux shows
a large jump at the beginning of RF injection, before the ion temperature begins to
rise. The tail temperature is 4.5 keV (see Fig. 6.2) at 0.8 s, when the ion temperature
is 3 keV. On this shot, the sight line of the PCX passes 0.21 meters below the center
of the plasma (r/a > 0.5 mapped to the midplane), where the ion temperature is
more than a factor of two lower than at the center.
The difference between the PCX Trail values for the bulk species and the central
ion temperature indicates that the tails observed by the PCX are not occurring in
the interior of the plasma.
6.2.3
Dependence on Edge Conditions
The ion tails' strong dependence on the edge conditions is revealed in two situations,
H-modes and toroidal field ramps.
One of the results of the plasma transitioning into H-mode is an immediate change
in the temperature and density profiles of the plasma in the scrape off layer[45] [46] [47].
.. ..
2.........
.....
Shot 960227042
N t F
..
4.
. . . .!...........
~ e~n M
i.........
.!...........
..
..
4..........
0o.po
070
o.
0o.po
0.90
iO
cm
t Cental
t . electron
PX
. glensit
a....
vn 1re ..........
signal.at.0.88
reais5t
wl is t .
du seconds
oin
......
,.....................
•
.-
.
4.--. . o.p
...
.
,
...........
0.p
0.70 T--e-ro-nin.
V
--------------------
o.po
ai
--..
---..
---.
. ---- -----------------.----
.. .
..........
. ........... :............. . ............
...-.
1;0
0.20
0.§0
0.70
0.00
in 10 cts/ms/keV
SNeutral D;Flux at--4.5 IeV
......... .......... . .. . . .
.. ....
3.4
......... ------------. . : ~ g t ss n ._. . !. . .... .::...........
. -. . --..
- -.-.-.-3.-
.......
. ---2 -'-----
.......:.....
:....
-------------------'--
...........:..........
o............. .................................'.
10
0.90
0.80
0.70
0.o0
Time inseconds
Figure 6.1: Edge heating during ICRF heating. Note PCX signal rises in4 ms and
remains constant, while the central ion temperature rises over 100 ms. Cutoff inPCX
signal at 0.88 seconds isdue to the PCX gate valve closing early.
32
r-lux auring IUhm
X
30
x 26 -
E
0)K
before ICRH
ICRH
24 -ore
a SFlux
Ttail from fit 4.5 ke
22
20
,,
0
5
, - ....
10
15
Energy in keV
20
25
Figure 6.2: Deuterium flux, In f(S), before and during edge heating by ICRF heating.
The gradients steepen, so that the temperature and density outside the LCFS drop.
The H-mode may be accompanied by fluctuations known as edge localized modes
(ELMs), which further modify the edge profiles.
Figure 6.3 shows an example of changes in the ion tails correlated with changes
in the plasma edge. The RF power turns on at 0.7 s, with only a small increase in
the flux of 5 keV neutrals. However, at 0.74 s the plasma transitions into H-mode, as
shown by the abrupt drop of the H, signal, and the neutral flux at 5 keV increases by
an order of magnitude. The neutral flux drops back down in less than 2 ms at 0.76 s
when the RF pulse turns off. It appears that the change in the edge parameters at
the H-mode transition change the susceptibility to ion tail generation. The ion tail
is still only observed while the RF power is on, however. The effect of an H-mode on
edge ion heating is further discussed in Sec. 6.8 of this chapter.
At low RF power, edge ion heating is sensitive to small changes in the toroidal
Shot 950201027
Neto F po5 erinMW
o.
3
.....................................
. .......................
T(..ro.).inkey
.
0.75
Neutral Flux at 5 keV in A.U.
0.70
300....-..
.
......................... .. ..
1..
.
.............. i
0.p0
.............
n AU...... ......
.---------- -- -----. ----..
. .--- --.---------------
-----------
o0.0
0.75
0.70
m: -----density
in 102o --------:Average electron
----------30 -- --------------------~-:i--
0.04.
.
.
0.70
----
-
-........0.75
----
0.80
Time in seconds
Figure 6.3: Change in neutral flux with H-mode
magnetic field. Figure 6.6 shows large changes in the energetic neutral flux with small
changes in the toroidal magnetic field. The peaks, which are reproducible, can have
FWHM as narrow as 1% in B 0 or less. As discussed more fully in Sec. 6.4, these peaks
in the energetic neutral flux occur when the RF frequency is at a cyclotron harmonic
in the edge. That is, when the magnetic field changes, the cyclotron frequency wei
also changes, and when wrf = wi in the SOL, large ion tails are observed by the
PCX.
This strong dependence on edge conditions suggests that the ion tails are located
in the plasma edge.
6.3
Edge Heating during H Minority Heating
Hydrogen minority heating at 5.3 T is the heating scheme most often used on Alcator
C-Mod. The high efficiency and the high powers available make it very attractive for
use on future machines and reactors[2]. Consequently, it is particularly important to
determine the prevalence and effects of edge ion heating in this regime.
Edge heating, as described above, is observed on some shots at this value of
toroidal field. We examined the data from 298 shots at 5.3 T with ICRF heating
for evidence of edge ion heating. These shots occurred between 12 January 1995 and
9 June 1995.
6.3.1
Scatter Plots
Scatter plots were used to search for thresholds and their possible parametric dependences. On the plots, each symbol represents a single shot, and indicates whether
edge heating was observed on that shot, as characterized by the fast jump in the
energetic neutral flux.
Several examples of scatter plots generated from the data are shown in Fig. 6.4.
' ''. .
. .
' ' I''
. . . . . II''
' '1 . . .
*
jump
0 no jump
Sjump
no jump
S>0
OOo
0000
0
*
0'
pup
, ,,, , I . .
10
0
r
)K
. ,,,
. . .. . I , ,
,, .. .
°*
0.0
40
30
20
D-port antenna voltage
*
jump
0
no jump
0.5
1.0
o-port antenna
0
15
power
*
0
*tump
)
O noump
OO
0
0
0
00
1.0
000
Il
. .. .I.........Iiii
0
10
20
. . I .........
30
max antenna voltage
I,,,,..
40
I....
50
...
0
10
20
30
max antenna voltage
Figure 6.4: Scatter plots of flux jumps at 5.3 T
40
50
The top two plots give information about the minimum threshold for edge heating.
On the top right plot, the axes show the RF power from each antenna during RF
heating. This plot shows that edge heating is only observed if RF power launched by
the E-port antenna is greater than 0.7 MW or D-port power is greater than 1.1 MW.
On the top left plot, the axes are the antenna voltage in kV. This shows a threshold
of 25 kV on the E-port antenna, or 20 kV on D-port. These plots show that edge ion
heating is not an effect of the total RF power injected into the plasma.
The relation between voltage and power is not the same for the two antennas
because of variations in their construction. It is interesting that the apparent power
threshold for edge heating is at very different power levels for the two antennas, but
at almost the same voltage. This strongly suggests that the edge heating is caused
by the electric field local to an antenna. The relationship between antenna voltage
and electric fields around the antenna is described in Sec. 2.2.3.
The term "threshold" may be misleading, because there is apparently not a firm
limit above which edge heating is always observed. It is reasonable to investigate
other plasma parameters that may have an effect on edge heating.
To be consistent with the conclusion that it's the maximum voltage that matters,
the other two scatter plots use the higher antenna voltage as an axis.
If the edge heating is purely a result of conditions at the plasma edge, then the
parameters to investigate are such things as edge temperature, density, impurities,
and so on. However, measurements of these quantities in the SOL are not available for
most shots. If we suspect that the edge heating is related to the core plasma, say via
poor central RF absorption, or that the relevant edge parameters are strongly linked
to core parameters, then searching for a parametric dependence of the RF voltage
threshold should prove fruitful.
The plasma parameters examined include plasma current, line-integrated density,
central electron temperature, gap between the LCFS and limiter,
Zeff,
and Ratiomatic
neutral density.
The two lower scatter plots show examples. None of the parameters examined
showed a clear effect on the voltage threshold.
Density, shown on the lower left
graph, shows a tendency for edge heating not to occur at values above 1.2 x 1020 m - 2 ,
but there are many shots below this density that do not show a jump in the flux.
Similarly, on the plot of maximum voltage versus outer gap, on the lower right, edge
heating appears to be more frequent at larger gaps. Neither of these conclusions is
very strong. Any effect of other parameters is even weaker than in the two cases
shown here.
The conclusion must be that edge heating depends on the antenna voltage and on
one or more other parameters poorly represented by the available measurements. We
believe that quantities such as density and temperature, even at the LCFS, should
be strongly related to the core values, but this is not true of density and temperature
in the outer SOL, near or outside the limiter radius. The poor correlation of edge
heating with core temperature and density may reflect its sensitivity to the plasma
parameters in the outer SOL.
6.4
Resonance at Plasma Edge
At certain values of the magnetic field the threshold for edge heating is much lower,
less than 10 kW. The low threshold is seen where there is an ion cyclotron harmonic
in the plasma edge, that is when wrf = eB/mi = fwi (integer f) with B evaluated
at the plasma edge. If the PCX sight line is horizontal through the midplane, then
the "edge" is near 0.91 m, outside the last closed flux surface.
Even though none of the standard C-Mod heating schemes places an ion cyclotron
resonance in the plasma edge, this regime is interesting for several reasons. Because
significant edge ion tails are generated at low RF power, this regime is ideal for
H, D flux at 4.85 keV during ramping BT
4
5
6
7
fRF/CD at edge
Figure 6.5: Hydrogen and deuterium flux during a TF ramp. Drop in D signal at
0.59 is due to a 50% drop in RF power.
studying the effects of the edge heating on RF coupling (see Sec. 6.7) and impurity
generation (see Sec. 6.6). Low-power RF shots are also more reproducible, making
this regime useful for making scans in PCX sight line and RF power.
A higher
percentage of RF power is deposited in the edge as well.
6.4.1
Toroidal Field Ramps
Figure 6.5 shows hydrogen and deuterium flux during a toroidal field ramp at constant
RF power. The relative magnitudes of the flux signals are correct.
During ICRF
injection, B ramps from 3.46 to 2.30 T, moving three deuterium cyclotron resonances
through the outboard edge of the plasma. In this sense, a cyclotron resonance occurs
at radius R if
£eB(R)
=
(6.1)
W= Wrj
mD
for integer f. The resonances therefore are defined with respect to the frequency of
the RF heating wave. For example, because the RF frequency on Alcator is 80 MHz,
the third cyclotron harmonic occurs where B(R) = 3.5 T. Because B(R) - BoRo/R,
the radial location of the third deuterium cyclotron harmonic is
R B0oRo
3.5T
Thus ramping the toroidal field strength has the effect of moving the radial locations
of the cyclotron resonances. At odd harmonics, that is when Eq. 6.1 is satisfied with
an odd integer 1, the deuterium signal is larger than the hydrogen signal, while the
hydrogen signal is larger at even harmonics. These peaks and the relative heights of
the hydrogen and deuterium flux are reproducible.
This shot also shows a small peak where wrf
= -wcH.
The cyclotron frequency
for a species depends on the charge-to-mass ratio, q/m, so this peak may correspond
to a cyclotron resonance of an impurity species, such as 06+. This peak and others
at fractional harmonic numbers are observed only occasionally.
The data in Fig. 6.5 was taken with the PCX analyzer horizontal through the
midplane. Figure 6.6 shows data taken during TF ramps with the PCX scanned to
various poloidal locations. The sight line for each shot is shown to the left, overlaying
the plasma and vacuum vessel surface. The structure of the signals, namely the
multiple peaks, was observed regularly.
In this form, the figure only indicates that away from the midplane, edge heating
occurs at lower B 0 . To put this in a more useful form, we calculate for each sight
1000-
D flux (4.6keV) during
--
500 ..................--
BT
ramp
------
0-
1000
500
-------
1000-0
400100
500
0-
"
200
50
0
4.2
-
--
4.4
---------
4.6
4.8
BTin T
Figure 6.6: Poloidal scan of PCX flux during TF ramp
5.0
line the radius of the resonance layer as a function of time. Because we want to know
this to accuracy of better than 1%, we need to include several terms in addition to
the basic 1/R dependency of the toroidal field. We correct for the TF ripple (as
described in Sec. 4.1.1) and the contribution of the poloidal field. Particles observed
at different sight lines in the poloidal scan do not have velocities purely perpendicular
to the magnetic field. As particles accelerated by RF waves at a cyclotron resonance
gain mainly perpendicular velocity, for each sight line we calculate the radius at which
the particles observed have purely perpendicular velocity, that is, Rb as described in
Sec. 4.1.2.
The result of this transformation is shown in Fig. 6.7. The bottom plot of the
stack of six panels shows flux surfaces, PCX sight lines, and the limiter. The dashed
lines are the flux surfaces out to the LCFS, and the dotted lines are flux surfaces
outside the LCFS. The heavy line represents the RF protection limiter, and the
thinner nearly-horizontal solid lines show the four sight lines used in this scan. The
x-axis is now radius of the resonance layer in meters. Note that in the bottom plot
the vertical axis has been compressed. These plots show clearly that the largest CX
flux for all the sight lines occurs when the resonance is just outside the limiter. This
is particularly clear in the case of the lowest sight line, where the limiter does not
follow a flux surface closely.
There are more hidden assumptions involved in making this plot. We have not
yet justified why the w f =
CWcD
surface is of any interest. Perhaps we should be
concerned with the radius where w f =
(£ +
0.1)wcD, or something else. Without
having presented any theoretical cause for the edge heating, the only justification is
our intuition that interactions between particles and RF waves will occur at cyclotron
resonances. These theoretical considerations will be presented later in Chapt. 8. As
partial justification, recall that the plots show that the flux jumps occur when the
resonance crosses the limiter radius to better than 0.5% in major radius. This con-
D flux (4.6keV) during BT ramp
1000
500
0
4
1000
50
0 ----------------- -------------
100
50 .
.
0
100,',,
o
....
-.....
F
50 ................... ..........................................
-0.30
,
0.75
0.80
0.85
R in m
0.90
0.95
Figure 6.7: Poloidal scan of PCX flux vs. resonance radius. Bottom panel shows flux
surfaces (dots and dashes), limiter (thick), and PCX sight lines (thin)
86
tinues to be true even at the lowest sight line where the spacing between the flux
surfaces (along the sight line) is significantly different and the limiter does not follow
a flux surface. This makes the connection extremely likely, though not proven.
6.4.2
Scans in RF Power
As mentioned above, shots with a resonance in the edge show large edge heating at
low RF power and are reproducible. With the toroidal field in the plasma center set
to 4.71 T, we performed scans in RF power during poloidal and toroidal scans of the
PCX sight line.
From the neutral flux as a function of energy thus obtained we calculate quantities
that have a more intuitive interpretation. The tail temperature is found as described
previously by fitting a line to In f between the energies of 1.5 and 3.5 keV. Estimates
of the power deposition during these scans are described in section 6.9.
The plots in Fig. 6.8 show the tail temperature at a given sight line as a function
of RF power, and Ttai as a function of sight line at a given power.
It is important to point out that the particles viewed on the various sight lines
have significantly different orbits. When the PCX scans toroidally, the actual point of
the plasma viewed changes only by a few cm, but the tips of the bounce orbits change
from being near the outboard midplane in front of the PCX analyzer to directly in
front of the RF antennas. Particles that have their bounce tips, for example, directly
in front of the antenna will cross the midplane at different toroidal locations with
different pitch angles.
So when we observe a change with toroidal angle, it may
represent a toroidal asymmetry more than a characteristic pitch angle dependency.
During toroidal scans, we observed significant ion tails only at pitch angles corresponding to banana orbits, not passing orbits.
2
>CLS1
1
8. 0
0.
E
.8
.6
0
4
1
>1 10
0
100
PRF in
TTal V. Zint:
1000
kW
PRF scan
4 --
4
2
-.....
c
--
-- 11kW
- -- 18kW
42kW
85kW
.506kW
8--.6-0
L
4000
005
010
015
020
0.-
-z(below midplane)
TTa V.PRF:
Toroidal scan
1
-
4-
>1
o
0
00
00
80
6
-
4
10
100
10
TTa v. Toroidal Angle: PRF scan
4 -
-
----
e
1
1kW
- -.18kW
-
42kW
2-
..-
85kW
-
-
. . . . . ....
...
0-
=1
80
0 6-.04
0
--
.
5
10
15
20
25
Toroidal angle
at LCFS
Figure 6.8: Ttal vs. RF power and sight line
88
30
3!
Shot 960119008
PCX flux at 3 keV
--- . . . . .. .. . ............------------6 0.-.................... ............. ............. ............... . . .
.....
2 0......
0
...............................
-.......
........
...... ............-
0.70
0.75
2
0.05
0.0
Net RF power in MW
. . -. .--. .2 .0 - - ------ -------. . .--
1- .........
0.0o
0.00
0.05
. . ..--. .
. .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .---.
.........
0.75
0.80
0.$5
0.00
0.D5
Time in seconds
Figure 6.9: PCX neutral flux and RF power with B 0 = 6.5 T
6.5
ICRF Heating at 6.5 T
Alcator C-Mod is operated with a toroidal field of 6.5 T when using mode-conversion
heating. While there is much less data in this regime, edge heating is observed with
a threshold approximately the same as that observed at 5.3 T. In the shot shown in
Fig. 6.9, the RF power was turned on and off quickly, but at varying power levels.
Edge heating occurs only when the power reaches above 500 kW.
6.6
Impurities
Impurities, referring to any non-hydrogenic species in the plasma, reduce central
plasma temperatures by line radiation and bremsstrahlung, can destabilize magnetic
equilibria, and dilute the density of hydrogen species[48]. Experiments on JT-60[41]
showed a strong correlation between edge heating and increases in impurity levels in
89
the plasma. The importance of impurities in tokamak operations and a possible link
between edge heating and impurity generation compel study of a similar possible link
on C-Mod.
6.6.1
Toroidal Field Ramps
In order to determine whether edge ion tails contribute to impurity generation during
RF heating, we examined shots which have high levels of edge heating at relatively
low RF power. This occurs when the RF frequency is an ion cyclotron harmonic in
the plasma edge.
When a resonance layer is moved through the plasma edge by ramping the toroidal
field, the flux of energetic neutrals increases by orders of magnitude. If either the
energetic ions in the plasma edge or the escaping neutrals caused impurities to be
sputtered or desorbed from the vessel walls, we would see an increase in impurities
at the same time.
Figure 6.10 compares two shots with similar levels of edge heating, but greatly
different RF power. The diagnostics used are described in Sec. 2.3.3.
The shot on the left has up to 2.7 MW of RF power injected into the plasma.
During RF heating, the percentage of radiated power increases, as do the levels of
molybdenum, showing that impurities are being generated. Note that on this shot
the plasma does not transition into H-mode, because the toroidal field is reversed.
During the shot shown on the right, the toroidal field is ramping, moving a cyclotron harmonic through the plasma edge around 0.65 seconds. The edge heating
is similar in magnitude (the PCX settings are identical) even though only 90 kW of
RF power is injected. There is clearly no change in radiated power or molybdenum
signals that would indicate that impurities are being generated.
960829023
960301045
e+14-....
A&Brightness
. ................
0.e
3.e ....M. Brightness . waveJength.,3 .....
wavelength.63.........
............
..
------------..............
.....
"A:
----------
..........
. - I ... .. . .......... .L
.e
..'l2-
.
.
0.0po
1o
060o
po
... M Brightness.-wavelength 128 ........
,e l& ...M6 Brightness.-.wavelength 128........
G.e+0& ......
.e+1....
opo........
oJ.
po
......--
opo
1;o
..
a-+.....
G.
0.0
opo
1;6
S....ABrightness waveJength.13........
& Brightness wavelength.13 .........
i
0.80--
- - -
- --
-
060
0.80
o
o po
-o po
10
4....... Main .plasma b6lometer in MW .........
-
MW .......
olomoee power.in
1s.........2P.b
......................
1-2..
; ...
e...
........
...
....... --------
.....
...
. ~.....
06-.-. -l
- ------400
----.......
----- CX
0.80
1:0
0.60
Piolometer
-1
powerin.MW ---------......... .,1
. =..... ........
""
k.,
0 ..........
n"
pnH-n080
' ......... 1 .........
. '1
........
metessiviy
.............
2 e.Q
.. 6.......
I.5e+12
.
---------.---
-
- . - ....... .............
-. .
-..
0.80
0.60
i=n,
1;o-
power.ia.V,
...... ,............
::2P.iblomete!
........
'6"
G.4
.-----.......
S
15e
.1.i..........
.
-ai
0
1 .....
0.5
0.9 .. ..
6.6 -
..........
---Z-meter-emissivity
L ..........
...
.............. ......... ..........
etE oer-in MW------Pi bo-.N
..........
02..
..........
..
0 . ...
.......
.. .. ............
I- k... 1...........
L..
. ..
1'0
080
0.60
..----. Z-meterni~ssivity
...
1'0
a.......... - Net RE poier.iq MW...........-2 ...... ...
,. .
.
....................
.....................
1.. ......... .. .......
9.-
.
4W00...............
PX
PC20.. . . --
.s
~
~ ........
4G..............~
...
..........
.5k. ............
.......
06 0 ..... 080
•
k--......
......... --
1V0
060
e -
Time in seconds
0
10
Time in seconds
Figure 6.10: Spectroscopic measurements during edge heating, at high and low RF
power, showing impurities uncorrelated with edge heating.
6.6.2
Results
These shots are characteristic of many examined, demonstrating that there is no
correlation between edge heating and impurity generation. That edge heating does
not contribute to impurities in Alcator C-Mod is in distinct contrast to measurements
made on JT-60[41], which showed that edge ion heating caused measurably raised
impurity levels.
6.7
RF Loading and Heating Efficiency
The shots described above are also ideal for looking at the effect of edge ion tails on
coupling the RF power to the plasma. The coupling of RF power from the antenna
to the plasma is believed to depend significantly on the characteristics of the edge
plasma[6], and so may deteriorate with edge ion heating.
The primary measure of the coupling is the loading resistance seen at the antenna.
Figure 6.11 shows traces for the E port antenna taken during a toroidal field ramp
at constant RF power. This shot is characteristic of many shots with toroidal field
ramps.
As the plot shows, the edge heating has no effect on the loading. This is an important negative result; the edge ion tails do not detrimentally affect the RF coupling.
6.8
H-modes
The H-mode is a regime of high confinement in which energy transport at the plasma
edge is reduced[45], and has been observed on many machines. A fusion reactor may
have to operate in a higher confinement regime such as H-mode to reach ignition.
The major indications that the plasma has entered H-mode are a sudden drop in D,
emission and changes in the slopes of the time traces of the plasma central density
Shot 960829033
0.2
...
0:2 I --
MW ............................
. . . . . .Net RF-.po . er.
. in.
0:2
0:2
0:
.. . . . *"
1
_
.............. ...... E-port RF ppwer.in -MW.
~o
0.......
------06.....
.....
0:2
-7::
.........
b ................ I................. ........
........
. ....
0:2 ..
0:2
1. --- .Voltag -onE- ort-antenna (peak.-ground) in kV ----------9:
87
-1
Jjjjl i.. . .. .
... .
.. .. ..
s
.f...
. ......
7..
o. . ..............
6.0 .................
iL6......
..................
............. RF,loading- of.E-Ipo antenna-in-Ohms ...............
2o
18 ---
ooo
o~yoi
: °...
'°...............
....
I..
......
........
....
,
...
..
,..,......
......P°.....
.o..........
..
14.
.......
---
keV
C X.fk.
-.....-.
...
1.b
...... ........... 1 6..................
.o..............0.10
10 ... .......... 6.66....
0
800:T
i...............
..
C...f~ ,Sk ............
..
........... ...................
......
.
.........................
.....
....
0.s
............... ................ ii6................. 1...................
,.e
Time inseconds
Figure 6.11: RF loading during edge heating.
93
and temperature.
MHD events called edge localized modes (ELMs) are sometimes observed during
all or part of an H-mode. ELMs increase the average transport across the LCFS, but
not up to L-mode levels[47][45].
6.8.1
Changes in Edge due to H-mode
Significant changes in the density and temperature profiles at the plasma edge as the
plasma enters H-mode are observed on many machines, including Alcator C-Mod[46].
Profiles outside the LCFS are measured with an array of fast-scanning Langmuir
probes[47].
When the plasma transitions to H-mode, density and temperature in-
crease inside the LCFS, while they decrease in the SOL. The changes in the density
and temperature profiles also affect the penetration of neutrals into the plasma.
The changes in edge heating with H-mode indicate that the edge heating is indeed
occurring in the SOL. Two different kinds of changes in edge heating are observed
with H-mode. On some shots, the flux increases sharply when the plasma enters an
ELM-free H-mode. On some other shots, the flux starts to increase slowly, with a
time scale of tens of milliseconds, peaking when the plasma enters H-mode.
6.8.2
Fast Jump at Transition
Figure 6.12 shows data from shot 960220007. The trace of D, light indicates that the
plasma enters an ELMy H-mode at 0.82 s and changes between ELMy and ELM-free
several times. The ELMy periods are seen on the D, signal as small increases in
emission with observable fluctuations. Looking at the deuterium flux at 1.8 keV, the
flux doubles when the RF turns on, but there is little effect on the flux at higher
energies.
When the plasma enters an ELM-free H-mode, the flux at 3.4 keV and
6.5 keV increases significantly. The fluxes drop rapidly when the H-mode becomes
ELMy.
Shot 960220007
3...
.......
Net BE.power.in.MW....
...
1:5.......Global .D emi~sion.in A.U.. ..
0..
0..
..
200............... PCX.D.fux.-.1..8.keV.. .......
400.-..........
PCX.D.flux.1:5 ----------- -- lba
ee i
34.keV
ini A . ..............
--------
PCX.D.flux..- 6.5.keV.. j ............
200...............
- - - --- - - . . ... .
----- - --- -. --e
0:0 ----------- -------
0.80
0.90
1I0
Figure 6.12: Edge heating during ELM-free H-modes
Shot 950512005
----- - - - - - 2 :0
-.
e R F p we r.i
-o
------........
............
kV..-D .. ----.C
1:0 ------- .........
C. ....
.............
----
5
0.00 -D,,erm
0.$5
0. OU -....
0. ......
5
6: S......
........0. !......
Global
issio-n
in-;
lll
.........
..
flu...
. ---------... keV....
.. ....
4:o
............
..
.
.f.
-.
.
1
.
.
.........
4 .-. .......
.... .. ....
,..
.... ...
..
......
.......-- -" --
5
0.,5.-.5 0.70
I o0.
-.
...
............
.C0....
flu.
.,......... o.,5.
: .......
2........ ...........C . D.fl. -.. .
....... ..........
Time inseconds
Figure 6.13: Precursor in PCX flux before H-mode
6.8.3
Slow Precursor
Figure 6.13 shows another type of behavior observed during H-modes. In this case,
the flux starts to rise after the RF power turns on, and increases on a time scale
of 10 ms. It is important to note that the PCX flux does not drop quickly, that is,
faster than 1 ms, during the notch in the RF at 0.67 s. This increase in flux is as
large as that observed with edge heating, but it changes with a significantly longer
time scale. Because the basic plasma parameters are changing on similar time scales,
we conclude that is a separate phenomenon from the edge heating discussed in the
earlier sections of this chapter.
96
6.8.4
Conclusions
Forming concrete conclusions is difficult because almost all strong C-Mod H-modes
are made using RF power. Differentiating between phenomena that result from Hmodes during RF heating and those that involve edge heating affected by the H-mode
is not possible on the basis of the data available. However, later we will show that the
changes in edge heating with H-mode are consistent with changes in the threshold for
parametric decay, which we will argue is the cause of the edge ion heating.
6.9
Power Deposition
Understanding the amount of the RF power that is deposited in the edge is obviously
important, for it tells us how much power is not available to heat the plasma center
and how much power is immediately reradiated as neutrals from the edge onto the
vessel surface.
6.9.1
Methods
It is straightforward to write down an equation for the amount of power in energetic
neutrals entering the PCX. If ci is the counting efficiency for channel i, ci is the
count rate, and Si is the energy of particles measured in the channel, then the power
in neutrals entering the PCX (in its energy range) is
PVC = YciS/ i
(6.2)
We subtract the count rate during an ohmic portion of the plasma to find the counts
due to the ion tail in the edge. Since there is no attenuation expected, the power
per unit area per unit solid angle leaving the tail in energetic neutrals at a particular
angle is
OPtaii,cx _
OQ0A
1
A
cii/
(6.3)
The proportion of particles lost from the tail by CX is simply the ratio of the
CX time to the total confinement time (tot,
the characteristic time for loss by all
mechanisms). With this, we can estimate the total power loss from the power loss
via CX. Thus we find the total power going into the tails is
OPtail,total
1
c i8 Tot,i
808A
A-'
CZ7cx,z
OA
it,(6.4)
A
Of course the confinement times are calculated as a function of energy (see Sec. 4.2).
The problem with this method is its dependence on several factors which cannot
be measured directly. In particular the power calculated above is proportional to the
term 1/(AQnoc), where c is an overall counting efficiency, as long as CX is not the
dominant loss term.
As described in Chapt. 3, it is straightforward to calculate the flux spectrum
from an ohmically-heated plasma given the parameters of a particular discharge.
Comparing data from the ohmic phase of a given discharge to this calculation, we
generate a scale factor, 0, which we apply to the data taken during RF heating.
Because the neutral density, no, in the SOL varies from shot to shot, this scale factor
is not a constant. Thus our final equation for the power lost to the edge ions is
8Poss
808A
1
=_
0(6.5)
A
cifi r'tot,i(65
ei
,i
To reiterate, this gives a calculated value for the power lost to the plasma edge during
edge heating which is independent of any errors in the absolute calibration of the PCX
or in the estimation of the neutral density in the SOL.
Power to ion tails
40 -.
30C
"Z
..
_
500 kW
20
20- -P
10
P = 80 kW
0
10
20
Pitch angle (from perp)
30
Figure 6.14: Power into edge ion tail as a function of pitch angle
6.9.2
Resonance in Plasma Edge
We have observed that the size of the edge ion tails at low RF power is greatest when
the RF is at a cyclotron resonance in the plasma edge. Because of this, we chose
this regime to measure the power going into the plasma edge. All other cases, in
particular hydrogen minority heating at 5.3 T, show a much lower percentage of RF
power being lost to the plasma edge.
In a series of plasmas with B 0 =--4.7 T, the PCX was used to measure the power
lost from the tails as a function of pitch angle, as shown in Fig. 6.14.
A central
magnetic field of 4.7 T puts the RF at the third deuterium cyclotron harmonic in
the midplane SOL.
This figure shows the results at RF power levels of 500 and
80 kW. Assuming the power is negligible at greater pitch angles, which is reasonable
considering how fast it falls off, we can integrate over pitch angle to get a total power
emitted per square meter. In this example, the power emitted is 10 kW/m 2
Power to ion tails vs. RF power
100 .00
. .
.
I
.
.
.'
+-....... + 12mm gap, perp
10.00
A 12mm gap, 20 0
A
C
._
cz
-
4mm gap, perp
1.00
°°'J
-~
0.01
"
10
100
1000
RF power in kW
Figure 6.15: Power into edge ion tail as a function of RF power
The total RF power going into edge ions depends on the extent of the edge heating. The narrow range in pitch angle of the emissivity in Fig. 6.14 indicates that the
suprathermal ions have bounce tips within 0.2 meters of the midplane. If the accelerated ions only exist near the antenna, say the distance from the PCX to the antenna,
then they occupy a toroidal extent of 1.4 meters, versus a total of 5.7 meters if the
edge heating is toroidally symmetric. Estimates of the plasma surface area with edge
2
heating are then 0.6 m 2 or 2.3 inm
, respectively. This leads to a total power lost to
the edge of 6 kW or 23 kW, that is, 1.2 or 4.6% of the total RF power. Because this
shot was chosen as one with large edge heating at low RF power, 4.6% power going
to the edge is the worst case. During high power ICRF minority heating on C-Mod,
the power lost to the edge via edge ion heating is always less than 1%.
Experiments at B0 = 3.5 T show how power to the edge depends on the RF
power. Data from three shots is shown. Two are identical, but with the PCX at
100
different sight lines. The power going to the edge ions at the given pitch angle varies
as (Prf)
4
. A shot with similar parameters, but with the gap between LCFS and the
limiter reduced, shows smaller edge heating, with the power to edge ions varying as
ePrf. A convective instability with growth rate proportional to RF power would have
an amplitude near threshold that varies as ePrf.
6.10
Summary
With the neutral particle analyzer, we observe large increases in energetic neutral
particle flux during ICRF heating. Because of the fast rise and decay times, the
energy dependence of the flux, and the dependence on edge parameters, we identify
this flux as being due to energetic ion tails in the plasma scrape-off layer.
Edge ion heating has no detrimental effects on the plasma. Spectroscopic measurements show no increase in radiation correlated with edge heating. The diagnostics
of the RF system show that the RF loading does not change due to edge heating.
Measuring the particle flux as a function of pitch angle, we calculate that the RF
power lost to the edge is always less than a few percent of the total power.
We have identified the dependence of the edge heating on various plasma parameters. During hydrogen minority heating, there is a threshold in RF electric field of
about 100 kV/m for edge heating, which changes during H-modes. When the ICRF
frequency is at an exact ion cyclotron harmonic in the plasma edge, the edge heating
is relatively large at low Prf (< 10 kW). These characteristics will be used in our
search for the mechanism of the edge heating.
101
102
Chapter 7
Mechanisms for Generating
Energetic Ions in the Plasma Edge
There exist a large number of processes by which the RF antenna can transmit power
to particles in the plasma edge[6][49]. This makes the topic simultaneously difficult
and interesting.
This work focuses on ions in the SOL with energies greater than a few keV. A
brief survey of the various edge-heating mechanisms will show that our observations
are consistent with heating by electrostatic modes.
Several well-known mechanisms of edge heating affect only the electrons. For
example, the motion of the electrons in the RF electric field can be randomized by
collisions with ions[49]. Also, the RF antenna can couple to global surface modes
with large electric fields that damp on electrons[15]. As they only effect electrons,
these mechanisms are not relevant to this work.
The paths by which power can be channeled into edge ions can be roughly divided
into three categories: direct damping of the electromagnetic wave, kinetic effects, and
generation of electrostatic modes.
103
7.1
Damping of the EM Wave
It is reasonable to ask if the ion tails observed when a harmonic is in the plasma edge
result from the fast wave damping on the directly on the ions.
In fact, this damping does not occur at any significant level. By looking at the
scaling of fast wave damping with temperature and harmonic number, it will be clear
that an accuracy of few orders of magnitude is sufficient.
The damping of the fast wave as a function of harmonic number n is proportional
to (k±p)2n/2nn!. It is also proportional to the radial width over which damping
occurs. Damping occurs when (w - nwci)/kllvti < 1, which translates to a radial
width of 6R = Rkll vti/nw.
These scalings lead to a strong temperature dependence
for high-harmonic damping.
Including the effect of harmonic number on wave polarization, the damping is[50]
2
Im(k)dx=2
Assuming typical edge parameters, n
n-1
w n
2 wpi R
7 c
1019 m -
4
3,
n - 1(7.1)
(n - 2)!
Ti = 10 keV, and harmonic number
of 6, the single pass absorption is 2 x 10- 24. The edge ion tails seen at Wrf = 6 WcD (as
shown in Fig. 6.5) cannot be the result of direct damping of the fast wave launched
by the RF antenna.
7.2
Kinetic Effects
Calculations of the ion orbits in the RF fields near the antenna have shown that these
can result in increased ion energies in the SOL.
104
7.2.1
Quiver Motion
Because the quiver velocity of the ions in the RF field is similar to the thermal
velocities in the edge, the average energy of the particles is increased[13]:
E = so +
q2 E2 (w
2m(w
2
-+w2w ) .
72.2
(7.2)
The electric fields near and between the blades of the Faraday screen on C-Mod
are around 400 kV/m at RF power levels where significant edge heating is observed.
In this case, the addition to the average ion energy is 6 eV. Quiver motion therefore
cannot be the source of the energetic ions observed.
7.2.2
Ponderomotive Force
The antenna RF electric field perpendicular to the magnetic field, particularly the
large field between the blades of the Faraday screen, exerts a ponderomotive force on
the plasma (see Appendix B),
2
q
-.
4m(w 2
-W)
-V
V |E
2
.
(7.3)
This leads to a drift velocity
V=
qB
qB
(7.4)
The drift moves ions outward in radius, and can cause them to hit the Faraday screen.
Assuming an electric field of 450 kV/m (between the blades) changing with a length
of 3.4 mm, the ion drift velocity due to the ponderomotive force is approximately
700 m/s. The additional particle energy due to the drift is much less than 1 eV.
Both quiver motion and ponderomotive drift result in higher ion energies in the
SOL when the RF is active. The energy difference, though, is only a few eV, and
105
hence much too small to be the source of the ions observed.
7.3
RF Sheaths
RF sheaths can result in ions with energies in the keV range hitting the vacuum vessel
walls and antenna components[51].
RF sheaths are caused by components in the RF electric field parallel to the
magnetic field lines on field lines that intersect a material surface. Electrons, much
lighter than ions, are accelerated into the surface, causing the DC plasma voltage to
increase. The resulting electric field is shielded by the plasma, leaving a sheath of a
few Debye lengths at the material. The DC voltage drop across the sheath is similar
in magnitude to the RF voltage on the antenna.
This effect could be the source of energetic ions that have been recorded with
probes. But as sheaths can only accelerate particles directly into a material surface,
they cannot be responsible for the ion tails observed on Alcator C-Mod.
7.4
Electrostatic Modes
The conclusion that electromagnetic waves from the antenna cannot cause the edge
ion heating observed in Alcator C-Mod may appear to also apply to electrostatic
waves. There are two crucial differences, however.
First, nonlinear processes can
generate electrostatic modes at lower frequencies, including oci, where damping is
strong.
Second, electrostatic waves can be excited with very large wave numbers
compared to the fast wave, which are also more strongly damped.
106
7.4.1
Parametric Decay Instabilities
The term "parametric decay" refers to a broad class of instabilities brought about by
coupling of several modes in a system.
Of interest here is the parametric decay instability by which the RF heating wave
decays into an ion Bernstein wave and an ion cyclotron quasimode[8]. The daughter
waves are electrostatic in nature, can have large wave number, and one or both has
a frequency which is an ion cyclotron harmonic. Parametric decay instabilities are
therefore a major candidate for edge ion heating, and will be explored in depth in
Chapter 8.
7.4.2
IBW Launching
There are several methods by which a fast wave antenna can excite ion Bernstein
waves(IBW) [15].
Measurements taken during IBW launching with a Faraday-shielded toroidal loop
antenna on Alcator C[52] and measurements of parasitic IBW excitation by a fastwave antenna on the ACT-I toroidal device[53] showed significant coupling to low k±
waves only. Theoretical predictions[54] also show that only low k 1 waves are launched
by an antenna. The condition of small k± corresponds to a situation where the RF
frequency is just below a cyclotron harmonic at the antenna (as shown in Fig. 8.1).
These waves would not damp on an edge plasma, particularly at higher harmonics,
because kzpi <K 1. Therefore, they cannot be a direct source of edge ion heating.
7.5
Conclusions
The ICRF can effect all the species in the plasma edge through a variety of mechanisms. They are all potentially important in antenna and tokamak design.
Our observations are consistent with one mechanism of edge ion heating, para107
metric decay instabilities (PDI). In the next chapter, theoretical calculations from
PDI theory are made and compared with experimental results.
108
Chapter 8
Parametric Decay Instabilities
The category of parametric decay instabilities (PDI) can include any system in which
a source of free energy and coupling between modes of different frequencies produce
oscillations at multiple frequencies that grow with time.
In plasmas heated by an electromagnetic wave, the heating wave can decay into
two or more modes at other frequencies. A wide variety of parametric decays are
possible, and many have been observed on tokamaks and other machines. Of interest
here are plasmas heated with a heating wave near the ion cyclotron frequency.
Parametric decays in the ion cyclotron range were first observed on the linear machine L-4 in 1977[55]. PDI was seen on the ACT-1 torus with fast wave heating[56].
Tokamaks recording observations of PDI during fast wave heating include ASDEX[39],
TEXTOR[40], and JT-60[41]. Experiments on DIII-D launching an IBW at a few
times the cyclotron frequency also showed PDI[42]. In these experiments, the identification of parametric decay was based on the frequency spectra observed with RF
probes.
In our data analysis, particularly in section 6.3.1, we searched for the onset of the
edge heating, a more tractable problem than understanding the magnitude of the edge
heating. Similarly, finding the point in parameter space where PDI becomes unstable
109
is more practical than attempting to calculate, for example, nonlinear saturation
levels.
The threshold for PDI will turn out to be approximately the same as the
thresholds observed for edge heating.
8.1
PDI Theory
In order to cause ion tails to appear in the edge, the parametric decay must put power
into a mode that can damp on the ions. This limits the kinds of PDI to examine.
Once an appropriate PDI type is identified, we must calculate the growth rate to
determine if decay of that type is unstable.
8.1.1
Waves
Of interest to us is the parametric decay in which the RF heating field, or "pump
wave", decays into an ion Bernstein wave and an ion cyclotron quasimode[8].
This PDI is an example of three-wave coupling, and there are two conservation
laws that must hold for the waves involved,
w 1 + w02 =
o0
and
1
+
2
=
0
.
(8.1)
As we shall show, the wavelength of the decay waves is much shorter than that of the
pump wave, basically allowing us to set k0 = 0. This is the dipole approximation,
and is in fact necessary for the theoretical treatment used. The frequency equation
implies that the frequency of the IBW will differ from the RF frequency by a multiple
of the cyclotron frequency. At 5.3 T, then WIBW = WRF - WcD
1.7WcD. If the RF is
at a cyclotron harmonic in the edge, then both the quasimode and the IBW will be
near a cyclotron harmonic.
110
8
6 ................................
6
........
... ........................
4.
..
............
----------------
---------------------
0
0.1
,
1.0
kpi
10.0
Figure 8.1: Typical dispersion relation for ion Bernstein wave, assuming n = 1 x
1019 m - 3 , T = 10 eV, k1l = 40 m - 1 , B = 4 T.
In principle, ion Bernstein waves propagate with kl = 0 or any kl satisfying[57]
W>3.
kllVte
The ion Bernstein waves involved in the parametric decay process, however, have
kl
$ 0.
Figure 8.1 illustrates a typical dispersion relation for ion Bernstein waves.
The ion cyclotron quasimode is not a propagating wave. It occurs near multiples
of wi at any k. Unlike propagating waves, the ion fluid current is in phase with the
electric field, so the power deposited, P = (E -f), is large.
111
8.1.2
Dispersion Relation
The derivation followed assumes that the plasma is infinite and uniform, and the
pump wave is treated in the dipole approximation. Collisions are neglected, as is the
magnetic component to the RF field.
An external electric field, F = (E0 d + E011
o ) cos w0 t, imposed upon a plasma in
a uniform static magnetic field B , will induce fluid motion
q
S2
S
Eoiwo
2 (cos wot
o-
WCe(
sinwot ) +
+-
E02
O
o
0
COS
Wot
(8.2)
where wc = qB/m for the species under question.
Using this to define an oscillating frame of reference, a modified linearized Vlasov
equation can be written for each species cr[58]:
0 F--+
at
& VxFo, + (&x WIcu)
VFo =
qo
qa
•1)
(VVfo)esinwa
sin Loot
-(1)
Vm,
(8.3)
In this equation, F and E (1) refer to perturbed quantities, while fo refers to the
unperturbed distribution function. Thus, this equation describes the propagation of a
wave through a plasma already disturbed by an external wave at w0 . The time-varying
term on the right hand side, p sin wot, couples waves with frequencies separated by
wo. This is the coupling between the ion quasimode and the IBW. The coupling term
/p,
is defined as
qu
=
a l
E11k1
E(j±.X
+
/o
r
1/2
2
+12
+LL0
(8.4)
(
w
2
c)
2
2
I
where a is simply an index representing electrons or one of the ion species present.
To proceed, it is assumed that t << 1. Physically, this means that the displacement
of a particle during an RF cycle is shorter than the wavelength of the decay wave.
112
This approximation is not necessary, but greatly simplifies the presentation of the
equations.
Solving the modified Vlasov equation (Eq. 8.3) with Maxwell's equations gives[8]
E(W1)
(W2) =
-
[La
-1
/t
12 [Xa(Wl) -
Xa(W2)][(Xv(W1) -
Xzv(W2))]"
(8.5)
The sums over v and a are over species in the plasma. The x's are defined as the
standard plasma susceptibilities, and other variables are defined in the customary
way; namely
1
Xa =
00
k2A
EFn(bo,)(1 +
(8.6)
ooZ((n)).
k2Do n=-oo
k2Vt,
ba =
w -
2w,
6(w)
As described in Eq. 8.1, kl
suppressed.
nwc
=Con
k jvtO
=
1
V2
= 2 T,/m
)7 X,(W)
-k 2 , so references to k in the above equations are
The function Z is the Fried-Conte plasma dispersion function, and
Fn(b) = In(b)e - b with In being the modified Bessel function.
The Debye length
of species a is AD. -
The relative velocity between different species, which appears in the term
[La-
/,
is the source of free energy that drives the instabilities.
Equation 8.5 has real and imaginary parts, and its complex solutions w 1 , W2 give
the growth rates of the instabilities.
113
8.1.3
Growth rate
In Eq. 8.5, we have on the left hand side the product of two susceptibilities, and on
the right, a term which we are formally considering as "small".
So we look for a
solution in which c(k, w2 ) _ 0 in order that the equation be satisfied.
Assuming the growth rate is much smaller than the frequency, we can make the
approximation
(W2) =
(w ± i)
(WR) + iOW z_ ER(WR) + CI(WR) ± i+y
(8.7)
The subscripts R, I refer to the real and imaginary parts.
For simplicity of notation, we now assume that only electrons and one ion species
are considered. The real and imaginary parts of Eq. 8.5 are then
1
1
12Re
CR(WR) " -l
4
Xi (WR))(Xe( W 1 )-
()(1)i
IC(W)
Xe(R))
(8.8)
CI
+7 ae((wR)
O4(R
I
1It
~
Pe 12 Im [1I
-
4
O+
6(wJ)
(Xi(w)
-
XS
(WR))(Xe (L 1 )-
Xe (WR))]
I
(8.9)
We solve for the growth rate
7
~-
-CIl
&0
+
pi
Pe
-
4-a_
2
Im
(i(w)
- X1)
i(R))(Xe(W1)
- Xe(WR))
.
8w
(8.10)
The solution to ER(wR) ~ 0 describes the ion Bernstein wave. So in Eq. 8.10, -ci/jR
describes the normal damping of the IBW. The term proportional to P2 is the PDI
driving term. It must be larger than the IBW damping term for the decay modes to
grow.
114
If (k, wi) do not describe a propagating wave then the imaginary part of the term
in the brackets will generally be "small", because E(wi1 ) is real and "large" over most
of its range. If w1 = Cwci, however, Xi( 1) is large and imaginary. This corresponds
to the ion cyclotron quasimode described above. At this frequency, then, we have
Xi(WR))
(Xi(W1) -
Xi()
(8.11)
.
Xi(wi)
E(Wl)
This leaves us with an expression for the growth rate of
7-
P ~- Pel2
-eI
E +
4
Im(Xe(wi)
IeV
- Xe (wR))
(8.12)
The condition on Xi leading to Eq. 8.11 is not always met, and we will keep the more
general expression when necessary in calculations.
8.1.4
Convection
If the IBW frequency is higher than the ion quasimode frequency, then there is always
a value for kii such that the growth rate given in Eq. 8.12 is positive. However, the
growth rate may be so small that the instability will not grow significantly on relevant
time scales.
A spectral component of the random noise spectrum of, say, density, characterized
by k in an RF field will grow in time as described in Eq. 8.10. At the same time,
it will move in space with the direction and speed of the ion Bernstein wave group
velocity. Because the RF field has a finite extent, the perturbation may propagate
out of the volume of interaction before it grows significantly.
If the spatial extent of the RF field along the group velocity, ig, is L, then the
time in the interaction volume is r = L/vg. If r7 < 1, then the wave is not unstable
in a convective sense. Traditionally, the threshold is taken to be at
T7
= 7r. If the
interaction region has dimensions (LX,Ly,Lz), then the time a perturbation remains
115
in the interaction region is
n
7 rmm
1 min
Lx,
LY
LZ
(VLm -X vL-y
v-z
In all calculations, these three terms were examined to find the proper interaction
time. For simplicity of notation, we will refer to the right-hand side above as L/vg.
The term
L/v, will be referred to here as the "PDI growth". The amplitude of
the decay wave, recall, is proportional to eL/v9g
.
The appropriate interaction volume to use must be entirely in the antenna near
field and have only small variation in parameters, including magnetic field. Because
of toroidal symmetry, the length in that direction is constrained only to be with in
the antenna near field, giving a length of about 0.5 m.
Radially, the dimension used is 10 mm. The LCFS of the plasma is 20-25 mm
from the front of the Faraday screen.
Density is approximately constant outside
6 mm from the LCFS, but electron temperature changes with a length scale of about
10 mm[10], so 10 mm is used as the radial width. The R dependence of B0 also
constrains the radial width. Looking ahead and using values for k1l and T at which
there is instability, and using a width in frequency for a resonance of
6w = kllVti
,
we find a radial width of 10-20 mm.
In the poloidal direction, the constraint is due to the curvature of the flux surfaces
and the limiters. A distance of 0.04 m above or below the midplane moves a flux
surface into a different magnetic field, as defined above.
Away from the midplane, the height of the interaction region would be BZ0.01 m.
BR
With an equation for the convective growth rate, it is possible to calculate the
threshold for instability as a function of various parameters in the range of interest
116
for C-Mod.
8.2
Numerical Calculations
In Chapt. 6, it was noted that having the RF at a cyclotron harmonic in the plasma
edge changes the edge heating dramatically. Parametric decay occurring during hydrogen minority heating would have an IBW frequency of 1.7wz, while PDI with a
resonance in the edge would generate an IBW at a frequency near fwci.
Because
the propagation of the IBW is very different in these two regimes, we treat them
separately.
All numerical calculations were done after separating the real and imaginary parts
of the equations as described above. All dispersion relations were solved for arbitrary
k. Terms of the order w/wce were neglected.
8.2.1
PDI during Hydrogen Minority Heating
Parameters and equations
During hydrogen minority heating, the RF frequency is at 2.7wi in the plasma edge at
the outboard midplane. Because the IBW frequency is not near a cyclotron harmonic,
the equation for the growth rate is simplified using
Im E(wn) - Im Xe(WR) = k2A 2
(8.13)
De
WR
where
= kVt e
We will soon consider the more general case, but if we assume that the quasimode
is exactly at wi, then
1
1.iBw = QM
117
Recall that k is the same for both modes. With this, we obtain a simple representation
for the growth rate,
J/F(
1
(-/e
=
2
aw kDe
At large (, e -
2
>
e - (1 .7)2,
2
c
-t2
2f
-
1. 7
-
)
4
e - (1.7) 2
(8.14)
,
meaning we can always choose a large enough C to make
the growth rate positive. At very large C, however, the growth rate will be positive
but very small, and the IBW will not be unstable in a convective sense. In practice,
convective instability is only found when 1 < C < 3.
As described in section 2.2.3, the largest component in the antenna electric field
is parallel to the current strap, Eo. For the decay modes,
--
implies
Eok
e
4
- 110
e
/~i - j
Mi
22
ci
2
02
2
while
|
k
Thus y is larger if k1
|.
implies
/pi-
pfe -
eEk
mi WOWci
Because the interaction region has different extent in
the r and 9 directions, we must always check both directions when searching for a
convectively unstable mode.
The Recipe
The method for applying these equations is comparatively straightforward. Plasma
parameters for the region of interest are chosen. For each k1j, the IBW dispersion
relation is solved to find k1 , and this is used to find vg for the wave and y. The
convective growth rate is then calculated for a range of kl to find the maximum. We
need only find a single kl that has -yr > r to show that parametric decay is unstable.
We vary the antenna electric field Eo, recalculating the maximum growth rate, to
118
PDI Threshold Electric Field vs. Electron Temperature
400
... .. .. .
Sme=2T
> 300 -- -----------...........
-c
-
Te=5Ti
Ti=3eV
S200
CD-
U_ 100rr--
0
10
30
20
Electron temperature in eV
40
Figure 8.2: RF electric field PDI threshold as a function of plasma temperature
find a threshold for PDI.
When the restriction that the quasimode occurs exactly at wc
is relaxed, the
procedure is the same as above, except there is an extra free parameter.
Results
Through the calculations, we found a threshold electric field as a function of density,
and ion and electron temperature.
Figure 8.2 shows the threshold in RF electric field as a function of plasma temperature. The density used is 1 x 1019 m - 3 . Most of this temperature dependence is
in the IBW group velocity, and not in the growth rate -y.
The most striking result is that the threshold for PDI in SOL type parameters is
very close to the experimentally observed threshold for edge ion heating. In the outer
SOL, where a temperature of a few eV is expected, the threshold is on the order of
100 kV/m (we will refine these parameters below), the same as the electric field in
the antenna near field at power levels around 0.5 MW.
119
50
PDI Threshold Electric Field vs. Electron Density
4 00 -
..
' '.i'I
I
.
I..
=1Te=15eV, T,=3eV
300
.\
.-
-----
Ve
Te=25eV, Ti=3eV
.
- -------
Te=25eV, Ti=6eV
= 200
100-
0100
u
....................
_-
)
1
7
9
1020
10
1018
101
1016
3
Electron
Figure
On
the
other
threshold
is
Figure
much
is
RF
electric
hand,
8.3
perature
at
higher
for
at
PDI
temperatures
than
the
both
ions
or
expected
effect
and
vs.
inside
RF
of
m
threshold
near
any
demonstrates
fixed
field
in
the
electric
plasma
as
density.
LCFS,
50
eV
or
so,
the
field.
density
electrons
plasma
shown.
on
the
The
threshold.
The
threshold
is
tem-
independent
3
of
density
outer
for
part
So
densities
of
far
the
in
frequency,
this
the
Figure
section,
eV,
=
ne
it
this
is
will
m-3.
At
plots
very
.
In
of
Te.
density
that
a
free
discharges
greater
the
quasimode
the
fast
on
Alcator,
the
1
is
ion
growth
calculations
Te,
than
at
x
1018
the
tails.
m-
ion
We
[47].
cyclotron
must
actually
parameter.
convective
These
low
L-mode
3
a
generate
as
of
m-
assumed
what
values
107
maintains
was
several
several
1019
x
frequency
shows
for
3
layer
quasimode
8.4
frequency
above
scrape-off
because
consider
3
8.3:
density
the
as
a
assume
convective
function
E
growth
0
of
=
is
large,
at
this
120
quasimode
kV/m,
but
Ti
is
=
peaked
2
at
wQM
=
there
would
high
enough
0.9
w,,.
not
for
In
be
the
this
any
mode
ion
case,
the
heating.
to
be
quasimode
At
would
higher
unstable,
Te,
and
120
the
grow
the
growth
peak
is
is
at
frequency,
lower,
wQM
and
though
=
wc%.
still
As
T.
T,,= 10eV
20
15
10
50 . ................. ........ ....
.........
0 1.....
2
0.70 0.80 0.90 1.00 1.10 1.20
0.70 0.80 0.90 1.00 1.10 1.20
O)QM/o)ci
.)QM Coci
Te= 15eV
T e = 20 eV
4.
2
1
1.20
1.10 1.20........
1.00 1.10
0.90 1.00
0.80 0.90
0.70 0.80
0.70
0.70 0.80 0.90 1.00 1.10 1.20
(OQM/O ci
Figure 8.4: PDI growth, yL/v 9 , as a function of quasimode frequency for T = 5, 10,
15, 20 eV. Threshold value, yL/vg = 7r is shown as a dashed line.
121
increases, the growth decreases, and the peak stays at the cyclotron frequency. So at
n = 1 x 1019 m - 3 , Ti = 3 eV, Eo = 120 keV, PDI with edge ion heating only occurs
for a range of Te, approximately between 12 and 18 eV.
In section 8.1.3 it was argued qualitatively that the highest growth rate should occur when the quasimode is near the cyclotron frequency. In the calculation above, we
found that the highest growth rate can occur with the quasimode significantly below
the cyclotron frequency at certain electron temperatures. The reason for this shift is
the strong dependence of the growth rate on the difference between the quasimode
and IBW frequencies.
H-modes
As described in section 6.2.3, the edge ion tails change sharply at H-mode transitions.
In the example shown in Fig. 6.3, we do not observe edge heating when the RF turns
on, but edge heating starts when the plasma transitions into an ELM-free H-mode.
The edge heating is large when the plasma is in ELM-free H-mode, and small in
ELMy H-mode or L-mode.
This effect can be explained by the PDI calculations above. During ELM-free
H-mode, energy transport through the LCFS is lower and the electron temperature
in the outer SOL is lower[47][45]. The low temperature reduces the PDI threshold.
With the onset of ELMs, energy transport to the SOL is increased and the electron
temperature rises quickly. This increase raises the electric field threshold, causing the
PDI and edge heating to stop.
8.2.2
Resonance in Plasma Edge
Parameters
If we examine a plasma in which the toroidal field value puts the RF at a cyclotron
harmonic in the plasma edge, the situation becomes more complicated. In the previ-
122
ous section, we found that the results of the calculations did not depend strongly on
magnetic field. In an IBW with WiBw ~ £- i, k_ is a strong function of WIBw, and the
wave has k±pi <K 1 just below a harmonic and k±pi > 1 just above a harmonic[57].
Also, if one of the modes involved in the decay is near an even harmonic of the
deuterium cyclotron frequency, then it is also near a hydrogen cyclotron frequency,
making it necessary to include the effect that the 0.5-15% of hydrogen in the edge
will have on the IBW propagation and damping. The equation for the growth rate,
Eq. 8.10, must then be expanded to include three species in the plasma. Terms proportional to any of ID - P/1,
/H -/1
IP,D - PHHI may be the primary driving terms
for the instability.
Growth Rate
We find that the calculated threshold for parametric decay is very low in this situation,
and the dominant coupling term is IPD - PHI -
It is possible to see this qualitatively.
field such that wRF
>
4
WcD
We choose, as an example, a toroidal
in the edge. The dominant decay is into a quasimode
at 2 WcD and an IBW with WIBW
> 2 WcD.
Because the IBW is near the cyclotron
harmonic, k 1 is very large. Hence, p is of order one, even when the mismatch in the
hydrogen and deuterium quiver velocities is small. Furthermore, when kl is small,
the IBW is essentially undamped. Also, the group velocity of the IBW is much lower
near a harmonic. So the coupling is large, the wave is undamped, and the time for
perturbations to transit the pump region is comparatively long. All these factors
contribute to the growth rate.
Results
Figure 8.5 shows the results of a calculation of the PDI growth, 7L/vg, versus magnetic field, simulating a Bo ramp.
Convective growth is shown for two decays:
123
PDI growth v.
300
200
100-
0
3.30
3.35
3.40
3.45
Bo in Tesla
Figure 8.5: PDI Growth, 'yL/v, vs. magnetic field.
B o =3.54 T.
124
3.50
4 WcD
3.55
in plasma edge when
WoQM =
cD,
IBW
3
WcD;
and WQM =
2
WcD, WIBW
-
2
WcD.
The peak of the higher
curve, at approximately 250, is much larger than the threshold value of 7r. The RF
electric field used, 10 kV/m, corresponds to an RF power level approximately 10 kW.
At the parameters used, the lower curve is actually not convectively unstable, but
would be at 20 kW of RF power.
A few caveats must be applied. First, the growth rate is unbounded as k --+ 0, so
the calculated growth is sensitive to the lower cutoff on k1l. We assumed that the kI1
of both decay modes must be greater than the kII of the launched fast wave. Second,
inclusion of a collision term would result in some damping of the IBW, reducing the
growth. Finally, because the IBW group velocity is relatively slow near a harmonic,
the transit time for a perturbation through the pump region is, in this example,
0.4 ms, which is slower than times for changes in the edge (in comparison, the transit
time calculated for B 0 = 5.3 T in the previous section is 1.5 ps). A more complete
calculation including all these factors must delve deeply into the physics of the plasma
edge, and is beyond the scope of the present work.
Nevertheless, given that the growth is a factor of 100 over the threshold at an RF
power of 10 kW, we expect that the parametric decay as described will occur.
Odd versus Even Harmonics
As shown in the traces in Fig. 6.5, hydrogen flux is high when the RF is at an even
cyclotron harmonic, and deuterium is large at odd harmonics.
This supports the conclusion that parametric decay is occurring due to the coupling term proportional to
I/D
-
IHl. For example, at
WRF
>
4
wcD
as above, PDI
can occur through electron-deuterium coupling into a deuterium cyclotron quasimode
and an IBW near 3 WcD, and would heat deuterium only. This turns out to be barely
unstable according to the calculation when Prf - 10 kW. But as we have just shown,
PDI with the quasimode at a hydrogen cyclotron frequency and an IBW near
125
2
WcD,
which will also damp on hydrogen, has a growth two orders of magnitude higher.
These large growth rates occur when the quasimode is at
1WcH.
At even deu-
terium harmonics, then, both of the decay modes will be at hydrogen harmonics, and
absorption on hydrogen dominates. At odd harmonics, the IBW will be at an odd
deuterium harmonic, and deuterium damping will be large. This is consistent with
the observations made with different harmonics in the plasma edge.
This argument that the behavior at different harmonics can be described by
changes in the growth rate must remain qualitative for two reasons: the saturation
level of the instability is not known, and the relative damping on hydrogen versus
deuterium has not been calculated. So far, we have assumed that the instability will
grow at least to the point where it has a measurable effect on the ions in the plasma
edge before its growth is stopped by some non-linear effect (Eq. 8.10 is valid only in
the limit that the amplitudes of the unstable waves is small). This saturation level
has not been calculated.
In general, it is well-known how to calculate the damping of a wave on different
ion species[30]. For electrostatic modes, such as the ion cyclotron quasimode and the
ion Bernstein wave, the ratio of damping on hydrogen to damping on deuterium is the
ratio of the imaginary parts of the linear susceptibilities (defined in Eq. 8.6), XH, and
XD. However, in calculating the damping in a parametric decay instability, currents
and electric fields must be calculated from Eq. 8.3, which includes the coupling of the
external RF wave to the decay waves.
8.3
RF Probes
The waves involved in parametric decay instabilities can be observed directly using
RF probes.
Although other experiments have reported correlations between PDI
spectra and fast particles in the edge[42][41][43][44], there is no reason to expect a
126
Spectrum from RF probe showing PDI
20
0
-
f
flBW
fICRF
- -20
C.
o -40
-60
-80
-80
0
, , , I ,
20
I ...
, I ,,
60
40
Frequency in MHz
80
100
Figure 8.6: RF spectrum from RF probe, showing a spectrum typical of PDI. B 0 is
5.3 T, and the deuterium cyclotron frequency is 30 MHz at the probe position.
strong correspondence between PDI-generated fast ions and PDI spectra detected by
the RF probes on Alcator C-Mod.
In some circumstances, RF probes on C-Mod have recorded spectra typical of
PDI, particularly at values of B0 that place a cyclotron resonance at the probe tip.
Figure 8.6 shows a spectrum taken at a central Bo of 5.3 T. The large peak at
80 MHz is due to the fast wave launched by the RF antennas. Interpreting the other
peaks as the results of PDI, we identify the peak near 30 MHz as an ion cyclotron
quasimode at the deuterium cyclotron frequency, and the peak at 50 MHz as an IBW.
8.3.1
Correspondence with PCX Measurements
Measurements from the RF probes, PCX edge heating measurements, and the theory
of PDI indicate that PDI measurements by the RF probes should be independent of
127
edge ion heating observed by the PCX.
The RF probes are positioned at different radii, and some can be scanned in
radius. A critical observation is that the quasimode peaks are at the frequency of the
cyclotron frequency at the probe tip, rather than the cyclotron frequency at, say, the
midplane LCFS[59]. It therefore appears that the oscillations measured with the RF
probes occur at the probes themselves. They are not waves generated elsewhere that
have propagated to the probes.
During hydrogen minority heating, the parametric decay has a threshold in electric
field at a value we expect only in the near field of the antennas, that is, within a few
centimeters. Therefore, the quasimode, which only exists in the pump region, would
only be measurable by a probe which is actually in the antenna near field and at the
correct radial location. The RF probes that provided the spectra shown above are
more than a meter away from the antenna.
The ion Bernstein wave should propagate out of the pump region. During hydrogen minority heating, though, we found that the IBW damping rate and the PDI
growth rate are of the same order. This implies that the IBW will damp away not
far from the antenna. A wave at 50 MHz has a vacuum wavelength of 6 m, too large
to propagate in vacuum regions inside the vessel.
8.3.2
RF Probe Measurements of PDI
As shown above in Fig. 8.6, RF spectra typical of PDI have been observed on C-Mod.
The data shown were taken during hydrogen minority heating, where the calculated
threshold for PDI is only exceeded near the antenna. The peak power in the decay
wave signal as picked up by the probe is 40 dB lower than the pump wave. This
does not account for variation in probe response due to frequency, wave number, or
electrostatic vs. electromagnetic waves.
However, there is a regime where the RF probe data are interesting, though we
128
Spectrum from RF probe showing PDI
40
20fd
v
2fci
3fd
4fci
fICRF
0-
a)
o -20
-40-
-60
,
0
_
,_,
20
,
,_
_..
_,,
40
60
Frequency in MHz
80
100
Figure 8.7: RF spectrum from RF probe, showing spectrum typical of PDI. Bo is
2.7 T, and deuterium cyclotron frequency is 16 MHz at probe position
do not expect a strong correlation with PCX data. As we have shown above, the
threshold in RF field for PDI is very low when wff = £wi,
= 2,3..., in which cases
the central RF absorption is also low. When central absorption is low, the RF fields
in the plasma edge are higher, high enough so that they are above the PDI threshold
all around the torus.
As described in Sec. 8.2.2, we calculate large growth rates for decay into a quasimode at WacH and an IBW near a multiple of WcD. The growth rate is large even at low
electric fields. For example, if w f = 5 wcD in the plasma edge, then the largest growth
is for decay into a quasimode at
cH
= 2 WcD and an IBW at 3WcD. A spectrum taken
with RF probes at such a field (Fig. 8.7) shows these peaks.
Considering the conclusions made above from the calculations above, there is no
reason to expect that parametric decay causing the observed edge ion tails would
be observable by the current set of RF probes. However, the spectra taken with a
129
cyclotron resonance at the probe support the conclusions from the calculations above.
8.4
Conclusions
In trying to explain the source of the edge ion heating, we must look for waves that
will damp strongly on the ions and drive tails in the direction perpendicular to the
magnetic field. Parametric decay into a large k mode at an ion cyclotron frequency
and another wave generates such a heavily damped mode.
Using parameters appropriate for the plasma in the outer SOL at the outboard
midplane during hydrogen minority heating, we found that PDI will be convectively
unstable at RF electric fields of approximately 100 kV/m. Experimentally, we observe
edge ion tails during hydrogen minority heating only if the voltage on at least one of
the RF antenna is above 20 kV, which corresponds to a near-field RF electric field of
about 100 kV/m.
The PDI threshold depends strongly on plasma temperature, and the experiments
showed large changes in edge heating with H-mode transitions, which cause rapid
changes in the temperature in the SOL. The fast change (< 1 ms) in the ion tails at
the time of the transition into ELM-free H-mode implies that the edge plasma itself
changes on the same time scale.
When the toroidal field is changed so that the RF frequency is at a deuterium
cyclotron harmonic in the plasma edge, the threshold for PDI drops to very low
values, well below 10 kV/m, which corresponds to an RF power of 10 kW. Similarly,
the experimental data show the largest energetic neutral flux when the RF is near
a harmonic in the edge. The PDI growth rates also predict that deuterium should
be heated if the third harmonic is in the edge, and predominantly hydrogen heated
if the fourth harmonic is in the edge. This is also reproduced by the PDI growth
calculations. Therefore, we conclude that the observed edge ion heating is explained
130
by parametric decay in the RF antenna near field.
131
132
Chapter 9
Summary and Conclusions
The Alcator C-Mod tokamak has proved to be a powerful and versatile machine upon
which to study edge ion heating due to ion cyclotron resonance frequency heating. In
particular, the good energy resolution of the neutral particle analyzer and its ability
to scan poloidally and toroidally allowed measurements that had not previously been
available in the study of this effect.
There were two goals of our study of edge ion heating on C-Mod. The first was to
determine what other, possibly detrimental, consequences the edge ion heating might
have on the plasma. The second was to to explain the source of the energetic edge
ions in terms of plasma physics.
9.1
Effects of Edge Heating
Experiments were performed to determine whether edge ion heating increased the
impurity level in the plasma, reduced the ability of the RF antenna to couple power
to the plasma, or served as a loss channel reducing the RF power reaching the center
of the plasma.
We chose a regime where the edge heating was large but the RF power level
was low, thus minimizing any effects of the RF not directly connected with edge ion
133
heating.
Indications of impurity rise were searched for in measurements of molybdenum
levels and radiated power. No changes in these were observed to correlate with edge
ion heating.
On the same shots, measurements of the RF system were examined for changes in
the RF coupling due to the edge heating. RF coupling, measured as loading resistance
of the plasma, was unaffected by edge heating.
To estimate the power being lost into the edge ions, we performed scans in pitch
angle to find the distribution of the fast ions. Combined with knowledge of the orbits
and confinement of these ions, we determined that the maximum RF power lost to
the edge is less than a few percent of the total injected RF power, and much less than
this in standard RF heating schemes.
We conclude that there are no detrimental effects to Alcator plasmas caused by
edge ion heating.
9.2
Parametric Decay Instability as Cause of Edge
Heating
Our broad data set includes measurements taken during dedicated experiments as
well as observations across the wide span in parameters at which Alcator is operated.
This gives us a broad base of observations to compare to known or expected causes
of edge heating.
Parametric decay instabilities, in particular decay of the RF heating wave into an
ion Bernstein wave and an ion cyclotron quasimode, can produce perpendicular ion
tails, resulting in parasitic power loss to the plasma edge.
We calculated the parametric decay growth rates at parameters corresponding to
the plasma edge during hydrogen minority heating to find the threshold for the decay.
134
The RF electric field threshold calculated for parametric decay is approximately the
same as the electric field near the antenna at the experimentally observed threshold for
edge ion heating. The decrease in the calculated threshold as electron temperature
decreases explains the increased edge heating observed during ELM-free H-modes.
This work is the first report of the effect of H-modes on edge ion heating by PDI.
Theoretically, parametric decay with large growth at very low RF power was
found when the RF frequency is near a cyclotron harmonic in the plasma edge. This
agrees with experimental results, where we found large edge heating at these magnetic
fields at low RF power. At odd harmonic numbers, the decay modes are expected
theoretically to damp strongly on deuterium, and damp strongly on hydrogen at even
harmonics, matching experimental observations.
Because of these correspondences, we are convinced that the edge ion heating
observed is due to parametric decay instabilities.
9.3
9.3.1
Comparison with Results of other Machines
Impurities
Experiments on JT-60 showed a correlation between parametric decay instability
with edge ion heating and increased impurity levels[41]. Specifically, the magnitude
of the IBW decay mode was correlated with the impurity radiation as parameters
were varied. The PDI was only observed with the antennas at (0,0) phasing.
On Alcator C-Mod, we have found that there is no increase in radiated power
and no other indication of increased levels of impurities associated with edge heating.
The simplest explanation of this apparent discrepancy may be the cliche "Correlation
does not imply causality." The plasma edge effects with (0,0) phasing are significantly
different for similar RF power levels at (0,r) phasing[40]. Certain conditions in the
edge, such as large poloidal electric fields, may be independently responsible for both
135
the impurity production and the PDI. The JT-60 results do not include an estimate
of the RF power going into edge heating. It is possible that, in their machine, a much
larger percentage of RF heating power went into edge heating. Greater power going
into the edge could result in greater impurity production.
9.3.2
Parametric Decay Instabilities
Results from ASDEX[39] included RF probe spectra and calculations of the PDI
threshold. PDI was observed with RF probes only at certain values of B0 , specifically
at fields such that
Wrf
=
WcD
p-4,5,5.5,6
in the edge. Their calculated growth rates were marginal for instability, 7L/vg ~ 1-3.
In Sec. 8.2.2 we found very large PDI in similar conditions by considering the modes
driven unstable by the ion-ion coupling (PD - PH).
This was neglected in their
analysis, and would provide high growth rates for the cases where a harmonic is in
the edge.
Conclusions from our experiments are supported by results of the parametric
decay experiments on DIII-D[42]. Parametric decay was observed only during limited
discharges, and not in diverted discharges. On DIII-D, diverted plasmas have a much
less dense edge (n, = 5 x
10
"5 m - 3 ) than limited plasmas (ne = 1 x 1018 m- 3 ). While
the ICRF heating in these experiments is 36 MHz, vs. 80 MHz on C-Mod, the trend
in Fig. 8.3, a large increase in threshold at lower densities, can explain why PDI
disappeared in diverted plasmas on DIII-D. This points out the importance of the
plasma density in estimating the effect of PDI on future machines.
136
9.3.3
Parasitic Power Loss
In plasma discharges on TEXTOR with and without PDI, the total RF power deposited directly into the edge is estimated to be at most 5% during ICRF heating[40].
Because the percentage is not affected by the presence of PDI, they concluded that
the power lost via PDI is significantly less than this.
The ion Bernstein wave heating experiments on DIII-D, however, showed approximately 50% of the RF power being deposited in ions in the plasma edge[42]. This
value was calculated from measurements made with a neutral particle analyzer similar
to the one on Alcator.
The C-Mod results are similar to those on TEXTOR, showing negligible amounts
of parasitic power loss.
9.4
Open Questions
Measurements of the fields of the pump and decay waves in front of the antenna
would make it tractable to study the relationship between the magnitudes of the
edge ion heating and the RF fields, and possibly connect the results to a theory
predicting saturation levels of the instability. Understanding the mechanisms setting
the saturation level of PDI is necessary to predict theoretically the effect of edge ion
heating.
The details of the damping of the quasimode on the ions, and the resulting changes
in the distribution function, should be calculated. The results could be compared to
our data, and would be useful in predicting the effects of edge heating in new regimes.
In the parametric decay growth calculations, effects of collisions, scattering, and
the three-dimensional geometry on propagation and damping of the ion Bernstein
wave should be included.
137
138
Appendix A
Magnitude of RF Electric Field in
Propagating Fast Wave
The magnitude of the RF electric field in the propagating wave can be estimated
using the the RF power flux, the wave number, and the polarization. We will use the
equations describing the fast wave to express the Poynting flux in terms of known
parameters and the electric field in the plasma.
The first step is to determine the perpendicular wave number of the fast wave.
, the dispersion relation for the fast wave can be
Assuming k = kA; + kz2, Bo0
written' [12]
A(1 - A)(h)
A - M1
(A.1)
/t) = , klS,B2 . 2
(A.2)
(A.2)
MI = A - M1
with the definitions
M2 --
2
V2
S(wr
I/kI l)2 w
rf
Omn
'Stix's unusually defined N1 , NII are here written as M±, M1
139
k-
VA
2
2
(Wr f /kI)
Wrf
B2
0omnj
2
(A.3)
and
2
c
A
W2
Ci
d
_ W2"
The value of kl for the launched wave is 11 m - 1 , as set by the structure of the antenna.
Using ne = 1020 m - 3 , B0 = 5.3 T at Ro, and kjI = 11 m -
1
in Eq. A.2, we find a value
for
M
4 x 10-4< 1.
Neglecting terms of order M ~, the dispersion relation, Eq. A.1, becomes
M_ = 1.
(A.4)
Using the above parameters in Eq. A.3 and solving Eq. A.4, ki _-65 m - 1 .
The next term needed is the polarization of the electric field, E, which is given
by[12]
Ex + iEy
Ex - iEy
_
We
, +++M
M
M2
c.
(A.5)
Neglecting M1 compared to unity, and rearranging terms
Ex
E-
iZrf
&j
(A.6)
The parallel component of the electric field, Ez, is shorted out by electron motion,
resulting in Ez < Ex[12].
Substituting for B from Faraday's law and neglecting Ez, the time averaged Poynt-
140
ing vector is given by
1
1
2po
2Io
ri
1
Wrf
(E kx + |E12kz
).
(A.7)
The surface area of each antenna is roughly 0.25 m2 . Using the values obtained
above for k and the polarization, Eq. A.7 allows us to calculate that the electric field
in the propagating part of the plasma is approximately 12 kV/m at an RF power of
500 kW.
141
142
Appendix B
Ponderomotive Force in a
Magnetized Plasma
Here we derive the ponderomotive force exerted on an ion in a radio-frequency electric
field in a magnetized plasma. Because we are concerned with the motion of ions in the
near field of a RF antenna, the RF component of the magnetic field is assumed to be
smaller than the electric field. For simplicity, we assume that E('o, t) = E, cos(wt)2
and B0 = B 0 . Thus the result is not completely general, but is applicable to the
situation of interest.
The standard derivation for the ponderomotive force in an
unmagnetized plasma is followed[60].
First, the electric field is expanded about the particle guiding center, so
E("
0
+ -, ,t)- E(ro) + (i 1 - V)E + . . .
(B.1)
The position of the particle is r = ro + r'l.
The Lorentz force law is split, using (F
V) as an order parameter (using "dots",
as in F, to refer to time derivatives), we find
mF1 = qE(Fo, t) + q(l
143
x
Bo0 )
(B.2)
as the first order in the expansion, and
mr2 =
q(F', - V)E(, t) + q(F2 x B 0 ) + q(rl x B 1 (Fo
0 , t))
(B.3)
as the second order equation.
The solution to Eq. B.2 is
_
rl =
(
-qE,
O2
-.
2
wC
-
COS(wOt)
.
W sin~wt)
wo
C
J
(B.4)
as can be easily verified by substitution into Eq. B.2.
This equation for r', and hence ri, can be substituted into Eq. B.3. Also, for
E(r, t) = E (r?) cos(wt), Faraday's law tells us that
-(#
Bl(F, t) - - (V
X V')
E )
sin (wt)
The drift due to the ponderomotive force occurs over many wave periods, T = 1/w,
so Eq. B.3 can be averaged over a period ((... )) to show only the drift motion. Only
terms proportional to cos 2 (Ot) or sin 2 (Wt) survive the averaging, as (cos(wt) sin(wt)) =
0. Upon averaging,
=-qEx
(1fi V)E(F-) cos(wt) )
OEx
2(.02- W) OX
(B.5)
and
(K
ri
OEx
-qEx
S
X(0,t)=
2
(W
2
_
o2)
9y
Dy.
(B.6)
These two terms have dimensions of force, and we define this as the ponderomotive
144
force,
-
2
-q2
4m(w 2 - W2)
-
(B.7)
2
Substituting in for the period-averaged term, we can rewrite Eq. B.3 as
m2
(B.8)
2Fp+q(
x Bo).
To get this, we must assume that the wave frequency and the cyclotron frequency are
somewhat separable. The solution to Eq. B.8 is
2
= p(cos(wct +
)
+ sin(wt +
as is easily shown by substitution.
145
)
q
B
t,
(B.9)
146
Appendix C
Near Field of RF Antenna
C.1
Vacuum Field
It is possible to calculate the near field of the antenna in the absence of plasma.
By definition, we will be interested in regions much less than a wavelength from the
antenna. The height of the antenna, about half a meter, is also much less than the
vacuum wavelength, which is 3.75 meters at 80 MHz.
Starting from Maxwell's equations,
0B
at
1 aE +
c at
and
V x B = 22
o,
(C.1)
we solve in terms of E and ' Assuming zero charge density in space, the solution is
V2, + WE2=
cE
-0
a
8
(C.2)
In the near field, scale lengths are small compared to the wavelength of a propagating
wave (by definition), so the second term on the left hand side may be neglected,
147
leaving
V 2 E =- 1
(C.3)
This may be written as the integral equation
E(xo) =
4 1
d
1|*- --oI
(C.4)
In the absence of a plasma, the current density is just that in the current strap,
and Eq. C.4 can be integrated numerically to find the electric field at any point in
space in the near field region.
Specifically, the profile shown in Fig. 2.8 was calculated using a simple model. The
current in the current strap was used with the correct 3-D shape, but feeders, the
vacuum vessel, and similar structures were neglected. The current strap was broken
into 200 objects poloidally and 20 toroidally. The second strap with the opposite
phasing is also included, though this had little effect on the profile in the region of
interest. The current density was assumed to be a constant in the strap.
C.2
Effect of the Plasma
In the presence of a plasma, the solution cannot be written in a simple integral form,
because there are currents in the plasma which must be consistent with the field. The
differential equation Eq. C.3 can be written approximately as
2
E
F
L2
I10Of
/o Ot
(C.5)
The additional term is the electric field due to the currents driven in the plasma.
148
In a cold unmagnetized plasma, the current is approximately[12]
.
3 =-CO W
(C.6)
Using this, we find that
L = c/Wpe
assuming w = w f = 2ir80 x 106 s-
2 mm,
1.
In a cold magnetized plasma, the RF current is not purely parallel to the electric
field. In the near field, however, the screening by the plasma is mostly due to the RF
plasma current driven parallel to F. Hence, for purposes of estimating the near field
when El
0,
2
3pi =
iWEO
2.
2
(C7)
In this case, the scale length is
L
W2)1/2
1/
=-
2
For n = 1 x 1019 m - 3 and B 0 of 5.3 T, L - 8 cm.
This term E/L
2
is the screening of the electric field by the plasma. Thus the
edge plasma screens out RF fields parallel to Bo very efficiently, but perpendicular
RF fields are unaffected on the relevant edge scale lengths.
To determine the near field of the RF antenna in the poloidal direction, we use
Eq. C.4 with no corrections for the effects of the plasma.
149
150
Bibliography
[1] CONN, R. W., Magnetic fusion reactors, in TELLER, E., editor, Fusion, volume
1, Part B, pp. 194-410, Academic Press, New York, 1981.
[2] PORKOLAB, M., Radio-frequency heating of magnetically confined plasma, in
TELLER, E., editor, Fusion, volume 1, Part B, pp. 151-193, Academic Press,
New York, 1981.
[3] KUNKEL, W. B., Neutral beam injection, in TELLER, E., editor, Fusion,
volume 1, Part B, pp. 103-150, Academic Press, New York, 1981.
[4] TAKASE, Y. et al., Plasma Phys. Control. Fusion 38 (1996) 2215.
[5] PERKINS, F. W., IEEE Transactions on Plasma Science Vol. PS-12, no. 2
(1984) 53.
[6] NOTERDAEME, J.-M. and VAN OOST, G., Plasma Phys. Control. Fusion 35
(1993) 1481.
[7] HUTCHINSON, I. H., Principles of Plasma Diagnostics, Cambridge University
Press, 1987.
[8] PORKOLAB, M., Fusion Eng. Des. 12 (1990) 93.
[9] FURTH, H. P., The tokamak, in TELLER, E., editor, Fusion, volume 1, Part
A, Academic Press, New York, 1981.
[10] LABOMBARD, B. et al., J. Nucl. Mater. 241-243 (1997) 149.
[11] LAO, L. L., ST. JOHN, H., STAMBAUGH, R. D., KELLMAN, A. G., and
PFIEFFER, W., Nucl. Fusion 25 (1985) 1611.
[12] STIX, T. H., Nucl. Fusion 15 (1975) 737.
[13] LEHRMAN, I. S. and COLESTOCK, P. L., Fusion Eng. Des. 12 (1990) 51.
[14] RYAN, P. M., ROTHE, K. E., WHELATON, J. H., and SHEPARD, T. D.,
Fusion Eng. Des. 12 (1990) 37.
[15] MURPHY, A. B., Fusion Eng. Des. 12 (1990) 79.
151
[16] HUTCHINSON, I. H. et al., Phys. Plasmas 1 (1994) 1511.
[17] IRBY, J. H. et al., Rev. Sci. Instrum. 59 (1988) 1568.
[18] MAY, M. L. et al., Rev. Sci. Instrum. 66(1) (1995) 561.
[19] GOETZ, J. A., LIPSCHULTZ, B., GRAF, M. A., et al., J. Nucl. Mater. 220-222
(1995) 971.
[20] NIEMCZEWSKI, A. P., Neutral particle dynamics in the Alcator C-Mod tokamak, PhD thesis, Massachusetts Institute of Technology, 1995.
[21] TOWNER, H. H. et al., Bull. Am. Phys. Soc. 26 (1981) 857.
[22] RICE, J. E., BOMBARDA, F., GRAF, M. A., MARMAR, E. S., and WANG,
Y., Rev. Sci. Instrum. 66(1) (1995) 752.
[23] RIVIERE, A. C., Nucl. Fusion 11 (1971) 363.
[24] ROQUEMORE, A. L. et al., Fusion Eng. Des. 12 (1990) 93.
[25] MILLER, J. C., Calibration and parametric study of the Alcator C-Mod charge
exchange neutral particle analyzers, Master's thesis, Massachusetts Institute of
Technology, 1995.
[26] COLCHIN, R. J., ROQUEMORE, A. L., and SCOTT, S. D., Rev. Sci. Instrum.
59 (1988) 1667.
[27] STILLERMAN, J. A., FREDIAN, T. W., KLARE, K. A., and MANDUCHI,
G., Rev. Sci. Instrum. 68 (1997) 939.
[28] GOLDSTON, R. J. and TOWNER, H. H., Journal of Plasma Physics 26 (1981)
283.
[29] GREENWALD, M., Personal Communication.
[30] HEIDBRINK, W. W. and SADLER, G. J., Nucl. Fusion 34 (1994) 535.
[31] TRUBNIKOV, B. A., Particle interactions in a fully ionized plasma, in LEONTOVICH, M. A., editor, Reviews of Plasma Physics, volume 1, p. 105, Consultants Bureau, 1965.
[32] BOIVIN, R., Personal Communication.
[33] KENNEL, C. F. and ENGELMANN, F., Phys. Fluids 9 (1966) 2377.
[34] BUZANKIN, V. V. et al., Magnetosonic-heating experiments in tokamak T-4
plasma, in Plasma Physics and Controlled Nuclear Fusion Research 1976 (Proc.
6th Int. Conf. Berchtesgaden, 1976), volume 3, p. 61, Vienna, 1977, IAEA.
[35] KAITA, R. et al., Nucl. Fusion 23 (1983) 1089.
152
[36] HAMMETT, G. W., Fast ion studies of ion cyclotron heating in the PLT tokamak, PhD thesis, Princeton University, 1986.
[37] NOTERDAEME, J.-M. et al., The role of the Faraday screen in ICRF antennae,
in European Conference on Controlled Fusion and Plasma Heating, volume 10C
part I, p. 137, European Physical Society, 1986.
[38] WESNER, F., PROZESKY, V. M., BEHRISCH, R., and STAUDENMAIER,
G., Fusion Eng. Des. 12 (1990) 193.
[39] VAN NIEUWENHOVE, R., VAN OOST, G., NOTERDAEME, J.-M., et al.,
Nucl. Fusion 28 (1988) 1603.
[40] VAN OOST, G. et al., Fusion Eng. Des. 12 (1990) 149.
[41] FUJII, T. et al., Fusion Eng. Des. 12 (1990) 139.
[42] PINSKER, R. I., PETTY, C. C., MAYBERRY, M. J., PORKOLAB, M., and
HEIDBRINK, W. W., Nucl. Fusion 33 (1993) 777.
[43] TAKASE, Y., PORKOLAB, M., SCHUSS, J. J., et al., Phys. Fluids 28 (1985)
983.
[44] PORKOLAB, M. et al., Phys. Rev. Lett. 38 (1977) 230.
[45] ASDEX Team, Nucl. Fusion 29 (1989) 1959.
[46] SNIPES, J. A. et al., Plasma Phys. Control. Fusion 38 (1996) 1127.
[47] LABOMBARD, B. et al., Phys. Plasmas 2 (1995) 2242.
[48] STANGEBY, P. C. and MCCRACKEN, G. M., Nucl. Fusion 30 (1990) 1225.
[49] COLESTOCK, P., GREENE, G. J., HOSEA, J. C., et al., Phys. Plasmas 12
(1990) 43.
[50] PORKOLAB, M., Plasma heating by fast magnetosonic waves in tokamaks,
in Advances in Plasma Physics, Thomas H. Stix Symposium, number 314 in
AIP Conference Proceedings, p. 99, Princeton, NJ, 1992, American Institute of
Physics.
[51] D'IPPOLITO, D. A. and MYRA, J. R., Phys. Plasmas 3 (1996) 420.
[52] TAKASE, Y. et al., Phys. Rev. Lett. 59 (1987) 1201.
[53] SKIFF, F., ONO, M., COLESTOCK, P., and WONG, K. L., Phys. Fluids 28
(1985) 2453.
[54] OKADA, H., ABE, H., ITATANI, R., and ONO, M., Phys. Fluids 29 (1986)
489.
153
[55] ONO, M., PORKOLAB, M., and CHANG, R. P. H., Phys. Rev. Lett. 38 (1977)
962.
[56] SKIFF, F., ONO, M., and WONG, K. L., Phys. Fluids 27 (1984) 2205.
[57] BERNSTEIN, I. B., Phys. Rev. 109 (1958) 10.
[58] PORKOLAB, M., Parametric instabilities and anomolous absorption and heating of plasmas, in Symposium on Plasma Heating and Injection, p. 46, Varenna,
Italy, 1972.
[59] REARDON, J. et al., Bul. Am. Phys. Soc. 41 (1996) 1549.
[60] CHEN, F. F., Introduction to Plasma Physics and Controlled Fusion, Plenum
Press, 1984.
154